HK40045493B - Media information processing method, device, electronic equipment and storage medium - Google Patents

Description

Media information processing methods, devices, electronic devices and storage media

Technical Field

This application relates to fields such as artificial intelligence, cloud technology, and big data, and in particular to a media information processing method, apparatus, electronic device, and storage medium.

Background Technology

In current internet media information, such as internet advertising, internet advertising can be divided into two types according to the different product forms: contract advertising and auction advertising.

The differences between contract advertising and auction-based advertising necessitate different placement strategies from media outlets. For contract advertising, the most important goal is to ensure sufficient ad volume—the number of ads placed must reach a predetermined target, neither too much nor too little. For auction-based advertising, the media's objective is to maximize platform revenue.

The different objectives of contract advertising and performance advertising necessitate different placement strategies. Currently, many media outlets can display both contract ads and auction ads on the same ad slot. Therefore, how to effectively display both contract ads and auction ads has become a pressing technical challenge.

Summary of the Invention

This application provides a media information processing method, apparatus, electronic device, and storage medium, which improves the rationality of playing media information by processing media information in a candidate media information set.

On one hand, embodiments of this application provide a media information processing method, the method comprising:

Obtain the target data of each first media information in the candidate media information set in the previous time period. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, playback competition information of media information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information.

For each of the aforementioned first media information, based on the target data of the aforementioned first media information in the previous time period, the first playback evaluation value of the aforementioned first media information in the current time period is determined;

Obtain the second playback evaluation value of each of the aforementioned second media information in the current time period, wherein, for the playback evaluation value corresponding to the aforementioned first media information or second media information, the playback evaluation value represents the probability of the media information being promoted;

Based on the first playback evaluation value corresponding to each of the aforementioned first media information and the second playback evaluation value corresponding to each of the aforementioned second media information, the media information to be played in the current time period is determined from the aforementioned candidate media information set.

On one hand, embodiments of this application provide a media information processing apparatus, the apparatus comprising:

The target data acquisition module is used to acquire target data for each first media information in the candidate media information set. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous period, media information playback competition information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information.

The playback evaluation value processing module is used to determine, for each of the first media information, the first playback evaluation value of the first media information in the current time period based on the target data of the first media information.

The aforementioned playback evaluation value processing module is used to obtain the second playback evaluation value of each of the aforementioned second media information, wherein, for the playback evaluation value corresponding to the aforementioned first media information or second media information, the playback evaluation value represents the probability of the media information being promoted;

The media information processing module is used to determine the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the first media information and the second playback evaluation value corresponding to each of the second media information.

In one optional embodiment, the aforementioned playback competition information includes at least one of the following pieces of information from each media in the aforementioned candidate media information set:

Click-through rate in the previous period;

Conversion rate in the previous period;

Exposure rate in the previous period;

The playback evaluation value of the previous time period.

In an optional embodiment, the playback evaluation value processing module described above is used for:

Based on the aforementioned target data, the historical playback evaluation value of the aforementioned first media information in the previous time period is adjusted to obtain the first playback evaluation value of the aforementioned first media information in the current time period.

In an optional embodiment, the above-mentioned media information processing module is further configured to:

Determine the probability of each of the aforementioned primary media information being played in the current time period;

The above-mentioned determination of the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the aforementioned first media information and the second playback evaluation value corresponding to each of the aforementioned second media information includes:

Based on the first playback evaluation value corresponding to each of the aforementioned first media information, the aforementioned probability of being played, and the second playback evaluation value corresponding to each of the aforementioned second media information, the media information to be played for the current time period is determined from the aforementioned candidate media information.

In an optional embodiment, the above-mentioned media information processing module is further configured to:

For each of the aforementioned first media information, obtain the display evaluation data corresponding to the current time period for the aforementioned first media information, wherein the aforementioned display evaluation data refers to information that affects the probability of the aforementioned first media information being played;

For each of the aforementioned first media information, the probability of the aforementioned first media information being played in the current time period is determined based on the display evaluation data corresponding to the aforementioned first media information.

In one optional embodiment, for any of the aforementioned first media information, the aforementioned display evaluation data includes at least one of the following:

Profile data of the target user corresponding to the current terminal device;

The device information of the current terminal device;

The attribute information of the aforementioned primary media information;

The current time information corresponding to the aforementioned terminal devices;

The current location information corresponding to the aforementioned terminal devices;

The above target users correspond to the behavioral statistics of the current terminal device.

In an optional embodiment, for each of the aforementioned first media information, determining the first playback evaluation value of the first media information in the current time period based on the target data of the first media information in the previous time period is achieved through a media information evaluation model, which is trained by a training module in the following manner:

Obtain a training sample set, wherein each training sample in the training sample set includes a sample target data of a first type of third media information in the initial time period;

The sample target data of each of the aforementioned third media information in the initial time period are input into the initial information evaluation model to obtain the predicted playback evaluation value of each of the aforementioned third media information in the next time period in the initial time period.

For each of the aforementioned third media information, based on the sample target data of the aforementioned third media information in the initial time period and the predicted playback evaluation value in the next time period, the first predicted evaluation effect characterization value corresponding to the aforementioned third media information is determined.

Based on the first prediction effect evaluation characterization value corresponding to each of the above-mentioned third media information, the first total training loss corresponding to the above information evaluation model is determined.

The information evaluation model is repeatedly trained based on the aforementioned training samples and the first total training loss until the total training loss meets the preset first training termination condition, thus obtaining the aforementioned media information evaluation model.

In an optional embodiment, each training sample further includes the actual playback evaluation value of the third media information in the next time period after the initial time period, the actual evaluation effect representation value in the next time period after the initial time period, and the sample target data of the first time period of the initial time period, wherein the first time period is the time period after the next time period. For each piece of third media information, based on the sample target data and predicted playback evaluation value of the third media information in the initial time period, the first predicted evaluation effect representation value corresponding to the third media information is determined through an effect evaluation model. The effect evaluation model is trained by the training module, which is used to:

The sample target data of each of the aforementioned third media information in the first time period are input into the initial information evaluation model to obtain the second predicted playback evaluation value of each of the aforementioned third media information in the first time period;

The sample target data of each of the above-mentioned third media information in the first period of the initial period and the corresponding second predicted playback evaluation value are input into the initial effect evaluation model to obtain the second predicted evaluation effect characterization value corresponding to each of the above-mentioned third media information.

For each of the aforementioned third media information, based on the aforementioned actual evaluation effect characterization value and the aforementioned second predicted evaluation effect characterization value, the first evaluation effect characterization value corresponding to the aforementioned third media information is determined. Based on the sample target data of the aforementioned third media information in the initial period and the actual playback evaluation value of the next period in the initial period, the second evaluation effect characterization value corresponding to the aforementioned third media information is obtained through the effect evaluation model.

Based on the first and second evaluation effect representation values corresponding to each of the above-mentioned third media information, the second total training loss corresponding to the above-mentioned effect evaluation model is determined.

Based on the aforementioned training samples and the aforementioned second total training loss, the performance evaluation model is repeatedly trained until the preset second training termination condition is met.

In an optional embodiment, each of the above-mentioned training samples further includes a playback effect evaluation parameter for at least one second-type fourth media information corresponding to the above-mentioned third media information in the next time period after the above-mentioned initial time period. For any of the above-mentioned third media information, the above-mentioned true evaluation effect representation value is obtained by the above-mentioned training module in the following manner:

Obtain the target play count of the aforementioned third-party media information and the play count of the next time period corresponding to the initial time period;

Based on the target playback volume and the playback volume, determine the playback effect evaluation parameters of the third media information.

Based on the playback effect evaluation parameters of the aforementioned third media information and the playback effect evaluation parameters of each of the corresponding fourth media information, the aforementioned true evaluation effect representation value is determined.

On one hand, embodiments of this application provide an electronic device, which includes a processor and a memory interconnected thereto; the memory is used to store a computer program; the processor is configured to execute a method provided by any possible implementation of the above-mentioned media information playback when the computer program is invoked.

On one hand, embodiments of this application provide a computer-readable storage medium storing a computer program that is executed by a processor to implement the method provided in any possible implementation of the media information processing method described above.

On one hand, embodiments of this application provide a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in any possible implementation of the media information processing method described above.

The beneficial effects of the embodiments of this application are as follows:

In this embodiment, target data for each first media information in the candidate media information set from the previous time period is obtained. This target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume from the previous time period, media information competition information, or target playback volume. For each first media information, a first playback evaluation value for the current time period can be determined based on the obtained target data from the previous time period. Then, combined with the obtained second playback evaluation values for each second media information in the current time period, the first and second media information in the candidate media information set can be sorted to determine the media information to be played in the current time period. Using this method, the first playback evaluation value for each first media information can be determined based on the target data, and the media information to be played in the current time period can be determined based on the first and second playback evaluation values. This effectively utilizes target data to determine the playback evaluation value of the first media information and uses the playback evaluation value to determine the media information to be played, improving the rationality of displaying the media information to be played.

Attached Figure Description

To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

Figure 1 is a schematic diagram of the framework of an advertising hybrid sorting system provided in an embodiment of this application;

Figure 2 is a schematic diagram of the application environment of a media information processing method provided in an embodiment of this application;

Figure 3 is a schematic flowchart of a media information processing method provided in an embodiment of this application;

Figure 4a is a schematic diagram of playing media information to be played through a target display interface according to an embodiment of this application;

Figure 4b is a schematic diagram of another method for playing media information to be played through a target display interface according to an embodiment of this application;

Figure 5 is a schematic diagram of the structure of an optional distributed system 100 applied to a blockchain system according to an embodiment of the present invention;

Figure 6 is a schematic diagram of an optional block structure provided in an embodiment of the present invention;

Figure 7 is a schematic diagram of the structure of a media information processing device provided in an embodiment of this application;

Figure 8 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

Detailed Implementation

The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

At least some of the media information processing methods provided in this application involve fields such as machine learning in the field of artificial intelligence, as well as various fields of cloud technology, such as cloud computing, cloud services, and related data computing and processing fields in the field of big data.

Artificial Intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating/interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning/deep learning.

Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

Cloud technology refers to a hosting technology that unifies hardware, software, network, and other resources within a wide area network (WAN) or local area network (LAN) to achieve data computation, storage, processing, and sharing. The media information processing method provided in this application embodiment can be implemented based on cloud computing.

Cloud computing refers to obtaining the required resources on demand and in an easily scalable manner through the network. It is the product of the development and integration of traditional computer and network technologies such as grid computing, distributed computing, parallel computing, utility computing, network storage technologies, virtualization, and load balancing.

Artificial intelligence cloud services are generally also known as AIaaS (AI as a Service). This is currently a mainstream service model for artificial intelligence platforms. Specifically, AIaaS platforms break down several common artificial intelligence services and provide them as independent or packaged services in the cloud, such as handling resource transformation requests.

Big data refers to data sets that cannot be captured, managed, and processed within a certain timeframe using conventional software tools. It represents massive, rapidly growing, and diverse information assets that require new processing models to achieve stronger decision-making, insightful discovery, and process optimization capabilities. With the advent of the cloud era, big data has attracted increasing attention. Big data requires specialized technologies to effectively implement the media information processing method provided in this embodiment. Technologies suitable for big data include massively parallel processing databases, data mining, distributed file systems, distributed databases, and cloud computing, among others.

This application does not limit the specific type of media information (first media information, second media information) involved. The media information (first media information, second media information) in this application embodiment can be media information formed by promoting products. Specifically, it can be applied in scenarios where there is a requirement to maintain the number of plays for product 1 (i.e., first media information) while maximizing the overall revenue of product 2 (i.e., second media information). The product can be a media product, such as a game product, a film and television product, a beauty product, a home furnishing product, an apparel product, or a daily necessities product, etc. Specifically, the product can be promoted through advertising, such as through contract advertising and bidding advertising, playing promotional content on smart devices, or playing promotional content on electronic display screens in shopping malls, electronic display screens on highways, etc.

The aforementioned smart devices include, but are not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. The specific devices can be determined based on the actual application scenario requirements and are not limited here.

To more clearly illustrate the technical solution of this application, this example uses a method where products are promoted and played through contract advertising and auction advertising, requiring a mixed sorting of contract advertising and auction advertising. Product 1 is played through contract advertising, and Product 2 is played through auction advertising. When playing contract advertising and auction advertising, the goal is to achieve the optimal overall balance between maximizing the guaranteed volume of contract advertising and maximizing the effective cost per thousand impressions (ecpm) of performance advertising. Here, contract advertising refers to the first type of media information mentioned above, and performance advertising refers to the second type of media information mentioned above.

In the current internet advertising market, based on the different product forms, internet advertising can be divided into two types: contract advertising and auction advertising.

Contract advertising, also known as display advertising, involves placing a predetermined number of ads at a predetermined price within a specified time period. It was the earliest form of online advertising sales. It refers to an agreement between a media outlet and an advertiser to place the advertiser's ads on certain ad slots within a specific timeframe. The settlement method is based on cost per time (Cost Per Time, CPT). Later, Guaranteed Delivery evolved, where the media outlet and advertiser agree to place a certain number of the advertiser's ads on certain ad slots for certain users within a specific timeframe. The settlement method is cost per mille (CPM). If the number of ads placed by the media outlet exceeds the advertiser's predetermined number, the excess will not be charged. However, if the number of ads placed by the media outlet is less than the advertiser's predetermined number, the media outlet will be required to provide financial compensation.

Pay-per-click (PPC) advertising, also known as performance-based advertising, refers to advertising where ads are selectively displayed based on the advertiser's bid. Advertisers set their own bids, and billing is based on performance. Common billing methods include cost per click (CPC) and cost per action (CPA). In recent years, optimized cost per click (OCPC) and optimized cost per action (OCPA) have emerged. Unlike contract advertising, performance-based advertising does not stipulate the number of ads to be displayed between media outlets and performance advertisers. Performance advertisers compete for media traffic through bidding, with the common traffic auction method being the Generalized Second-Price (GSP) mechanism.

The differences between contract advertising and auction-based advertising necessitate different placement strategies from media outlets. For contract advertising, the most important goal is to ensure sufficient ad volume—the number of ads placed must reach a predetermined target, neither too much nor too little—while simultaneously considering metrics such as click-through rate (CTR) and conversion rate (CVR). For performance-based advertising, the media's objective is to maximize platform revenue, specifically, to maximize overall eCPM.

The different objectives of contract advertising and performance advertising necessitate different placement strategies. Currently, many media outlets can display both contract ads and performance-based ads on the same ad slot. Therefore, ad placement systems often incorporate a dedicated module—a mixed sorting of contract/performance ads—to determine whether to display contract/performance ads and which type of contract/performance ad to show.

The terms used in this application are explained as follows:

Advertisers: People who want to advertise their own brand or products.

Media (publishers): The platforms that provide advertising space. Examples include news apps, browsers, short video platforms, film and television platforms, instant messaging applications, etc.

Advertisers (agency): In essence, they are intermediaries who help advertisers find media advertising space and help media find advertisers.

Audience: People who "consume" the advertisement, i.e., consumers or users.

Click-through rate (CTR) is a common term in internet advertising, referring to the click-through rate of online ads (image ads/text ads/keyword ads/ranking ads/video ads, etc.), which is the actual number of clicks on the ad divided by the number of ad impressions.

Click-value rate (CVR) is a metric for measuring the effectiveness of CPA advertising. Simply put, it's the conversion rate from a user clicking on an ad to becoming a valid activated, registered, or even paying user. CVR = (Conversions / Clicks) * 100%. Conversions need a clear definition, which varies depending on the individual. Generally, it refers to the standard used by the advertising industry client (usually the advertiser) to evaluate the agency (usually the distribution channel). If this standard is a valid mobile phone number registered user, then the conversion rate (the conversion rate from click to valid user) is: CVR = Number of mobile phone number registered users brought in by the channel / Number of clicks brought in by the channel. Generally, 0 ≤ CVR ≤ 100%.

Predicting click-through rate (PCTR) refers to estimating the probability of an ad being clicked before it is displayed under certain circumstances.

Effective cost per mile (ecpm) is a metric used by media outlets to measure the return on their advertising campaigns. It refers to the revenue generated for each thousand ad impressions. A higher ecpm is generally better for media outlets. ecpm = CPC × CTR × 1000.

Cost Per Mille (cpm) is the cost per thousand ad impressions. Because the cost per impression is very small, the industry standard is to charge based on 1,000 ad impressions. Ads charged this way are mostly for brand promotion and product launches.

Billing is based on the number of times an ad is clicked (Cost Per Click, or CPC for short). Keyword bidding and feed ads mostly use this model.

Optimized Cost per Click (OCPC) is implemented, charging based on CPC. The more accurate and scientific conversion rate prediction mechanism helps advertisers acquire more high-quality traffic while improving conversion rates. Based on the advertiser's bid, the system intelligently and dynamically adjusts bids according to multi-dimensional, real-time feedback, and massive amounts of historical data, as well as the predicted conversion rate and competitive environment. This optimizes ad ranking, helping advertisers win the most suitable traffic and reduce conversion costs.

Billing is based on user feedback and actions (cost per action, CPA), typically for registration, which incurs registration costs. It also includes the cost per download (Cost Per Download, CPD) and the cost per installation (Cost Per Install, CPI).

Optimized Cost per Action (OCPA) is essentially still a CPA-based payment system. When advertisers select specific optimization goals (e.g., mobile app activation, website order placement) during the ad campaign process, provide the average price they are willing to pay for these goals, and promptly and accurately report performance data, we will use a conversion prediction model to estimate the conversion value of each click for the advertiser in real time, automatically set bids, and ultimately charge per click. Simultaneously, our conversion prediction model will continuously and automatically optimize based on the advertiser's ad conversion data.

Cost per sell (CPS) is a direct performance-based marketing advertising method.

Cost Per Time (CPT) is the payment method for ad placements, typically for one day, one week, or one month. Paying a fixed price to display ad space for a specific period is considered the most hassle-free advertising method. For most platforms, CPT is the fastest and most effective way to generate revenue.

Traffic can be understood as PV, which stands for Page Views or Clicks, and it measures the number of times a website or webpage is visited by users. Specifically, the PV value is the number of pages on a website or the number of times a specific webpage is viewed by all visitors within 24 hours (0:00 to 24:00). PV refers to the number of times a page is refreshed; each page refresh counts as one PV. The measurement method involves the browser sending a request to the web server. Upon receiving this request, the web server sends the corresponding webpage to the browser, thus generating one PV. Therefore, as long as the request is sent to the browser, regardless of whether the page is fully loaded (downloaded), it should be counted as one PV.

To achieve a hybrid ranking of contract ads and performance ads, this application example primarily employs game theory and reinforcement learning to utilize the guaranteed volume of contract ads, the eCPM distribution of performance ads, and the inventory distribution of contract ads (i.e., unfinished plays) to determine a reasonable bid for contract ads, thereby achieving the optimal goal of maximizing both the guaranteed volume of contract ads and the overall eCPM of performance ads.

As an optional implementation, Figure 1 shows a schematic diagram of the framework structure of an advertising hybrid ranking system provided in this application. As shown in Figure 1, the system includes both offline and online components. The offline component mainly performs model training and traffic prediction. The online component mainly uses the model prediction function to provide the bid for contracted advertisements based on real-time online fine-tuning and exposure.

As shown in Figure 1, the offline part of the ad hybrid ranking system may include a traffic prediction module, a simulator, and a model training module, while the online part may include a model prediction module, an ad hybrid ranking module, and a user terminal.

Taking a specific day (i.e., a one-day period) as an example, the process of sorting contract ads and performance ads in a mixed manner and then playing them is as follows:

Step S1: Divide the campaign period of the contract advertisement into fine-grained time slices (e.g., 5 min, 10 min, 15 min, etc.).

Step S2: At the beginning of each time slot, based on the playback status of the previous time slot (i.e., the historical playback volume of the previous time period), the competition situation in the mixed playback stage (i.e., the aforementioned media information playback competition information), and the traffic prediction information of the contract advertisement, reinforcement learning (RL) is used to determine the bid for the contract advertisement. The bid is the aforementioned playback evaluation value; the bid can also be understood as scoring the contract advertisement to obtain the final probability of the contract advertisement being played.

Step S3: Observe a time slot and record relevant information generated in the previous time slot when competing based on the bid of the previous time slot, such as the playback status of the previous time slot and the competition status during the mixed play phase.

Step S4: Repeat steps S2-S3 until the contracted advertisement plays to the predetermined number or the campaign period ends.

It should be noted that when sorting contract ads and performance ads in a mixed manner, the number and specific contract ads to be played within a given day are known in advance. The performance ads to be played will be scheduled to play different time slots based on the current playback situation.

To more clearly explain the function of each module, the following is a separate introduction to each module:

The traffic prediction module is mainly responsible for estimating the traffic of contract ads. Its predictions include the traffic distribution of the contract ad over time, the competition for the contract ad on the corresponding traffic (including the number of contract ads, CTR distribution, etc.), and the competition for performance ads on the corresponding traffic (including the number of performance ads, ECPM distribution, etc.).

When the traffic prediction module estimates the traffic distribution of contracted ads over time, it can use a traffic prediction algorithm. This algorithm predicts the traffic distribution of a predetermined audience within a predetermined region based on historical traffic data of the app. For example, traffic distribution prediction means that if a contracted ad is targeted at male traffic in region B of a video app A from January 2nd to January 4th, 2020, the algorithm will estimate the specific amount of male traffic in region B of video app A from January 2nd to January 4th, 2020, based on the daily distribution of male traffic in region B on video app A over a past period.

Typical traffic prediction methods generally employ machine learning models, and the prediction process is as follows:

Step 1: Collect the advertiser's targeting criteria and calculate the daily traffic volume under those targeting criteria;

The targeting conditions can be set according to the actual needs of the scenario, and are not limited here. For example, the targeting conditions can be for traffic in certain specific regions (such as first-tier cities), traffic for certain specific users (such as China Telecom users, China Mobile users, China Unicom users, etc.), traffic for specific groups of people (such as people aged 20-40), and so on.

Step 2: Use the advertiser's targeting conditions as features and the corresponding traffic volume as labels to train a machine learning model;

Step 3: For a new advertiser who places an order, use the machine learning model trained in Step 2 to predict their traffic distribution during the campaign period.

The simulator primarily simulates the mixed-ranking and display phases of advertising, capable of mimicking the realistic mixed-ranking process. It comprises two sub-modules: a mixed-ranking simulation module, which simulates the online logic of the mixed-ranking phase (including ranking rules, maximum number/duration limits for playable ads per ad slot, and filtering logic for ads from the same advertiser); and a display simulation module, which is a statistical machine learning model. The model's input features include user characteristics (age, gender, education level, interests, shopping behavior data, recently viewed browsing behavior data, etc.), ad attributes (ad ID, industry, duration, etc.), context (time, location, network, user device, etc.), and user behavioral statistics (number of videos watched that day, number of ads viewed, etc.). The output of the display simulation module is the probability that an ad will be played.

The online logic process of simulating the mixed-sorting stage can be implemented through the algorithm of the mixed-sorting simulation module. The specific inputs and outputs of the algorithm of the mixed-sorting simulation module are as follows:

Input: the ranking log of the ad placements in a 5-minute time slice (such as the target data of the first media information in the previous time slice), the contract ad A, and the new bid new_bid (such as the first playback evaluation value) of the ad (such as the first media information) given by the reinforcement learning model (i.e. the media information evaluation model).

Among them, embedded advertising refers to a communication platform that uses media such as Compact Disc (CD), Video Compact Disc (VCD), Digital Video Disc (DVD) or packaging posters to convey brand and product information to target consumers in a short period of time through distribution agencies covering the whole country and some overseas markets. It is also called "advertising with the film".

Output: The number of times the contract ad won during this 5-minute time slice, and information such as the actual competition situation.

The specific algorithm for the mixed-row simulation module is as follows:

Step 1: Collect the sorting traffic X of contract ad A during the 5-minute time slice;

Step 2, for each traffic x in the sorted traffic X:

a. Calculate the new score new_score of contract ad A on each traffic x in X. Generally, new_score = new_bid * pctr * pcvr;

b. Initialize the winning ad queue Res to be empty.

c. For each ad in a pre-roll ad:

1) Sort the ads that can be displayed on this post from highest to lowest score, and mark the ad queue as L.

2) For each ad l in the ad queue L, if the total duration of ad l and the ads in the winning ad queue Res does not exceed the total duration allowed for the ad slot, and ad l and the ads in the winning ad queue Res are not from the same advertiser, then add ad l to the winning ad queue Res; otherwise, continue execution.

d. Determine whether ad A is in the winning ad queue Res, record whether it wins, and the average score of the competing queue, etc.

Step 3: Summarize the data and provide relevant statistics such as the win rate of Ad A and the competitive environment.

The winning ad queue Res mentioned above represents the queue of ads that can ultimately be played.

Specifically, assuming a video app's intro ad slot can play eight pre-roll ads with a total duration of 150 seconds, the eight pre-roll ads to be played can be determined as follows:

The specific duration of each ad can be preset. For any given post, the ads that can be displayed on this post are sorted from highest to lowest score, and the ad queue is marked as L. The ad with the highest score is selected, such as ad 1. If the total duration of ad 1 and the ads in the winning ad queue Res does not exceed the total duration allowed for the ad slot, and ad 1 and the ads in the winning ad queue Res are not from the same advertiser, then ad 1 can be included as an ad in the winning ad queue Res.

Model Training Module: Contract advertising is a typical reward delay problem. The true performance of the contract advertisement can only be known after the advertisement's playback period ends. This problem can be solved using reinforcement learning modeling. Reinforcement learning models can include Reinforcement Learning (RL), Deep Q Network (DQN), Deep Deterministic Policy Gradient (DDPG), and Actor-Critic Algorithm (A3C), among others. No specific model is limited here.

The specific design of State, Action, and Reward in reinforcement learning is as follows:

State: The state consists of three parts: the competitive environment state during the mixed ranking phase, the traffic prediction state, and the delivery state. The competitive environment state during the mixed ranking phase can include the number of contract ads participating in the mixed ranking, the bid distribution of performance ads, the bid distribution of other contract ads, win rates, CTR/CVR distributions, etc. The traffic prediction state can include the predicted inventory distribution corresponding to the contract ad, the ECPM distribution of performance ads on the corresponding traffic, and the distribution of the number of competing contract ads, etc. The delivery state can include the pre-ordered volume of contract ads, the number of ads already exposed, etc. The target data of the first media information in the previous time period and the sample target data of the third media information in the initial time period mentioned above, as well as the specific information contained in the initial and final states mentioned later, can be found in the specific information within that state.

Action: The bid for the contracted advertisement. This action is the predicted playback evaluation value described above.

Reward: The reward is a piecewise function, mainly consisting of three parts. The calculation method for the actual evaluation effect representation value of the third-media information in the next time period after the initial time period, as mentioned earlier, can be referenced in this reward calculation method.

In each time slot (such as the next time slot after the initial time slot mentioned above), if the total number of completed plays for the contracted advertisement up to the current time slot has not yet reached the predetermined number (such as the target number of plays for third-party media messages):

Therefore, Reward = Contract Ad Views + Alpha * Contract Ad Performance + Beta * Performance Ad ECPM.

If the number of views for the contracted ads has reached the predetermined target:

Therefore, Reward = - Contract ad views + alpha * Contract ad performance + beta * Performance ad eCPM.

In the above formula, the number of contract advertisement plays can correspond to the number of plays of the third media information in the next period after the initial period, the contract advertisement effect can correspond to the playback effect evaluation parameter of the third media information in the next period after the initial period, and the performance advertisement ecpm can correspond to the playback effect evaluation parameter of the fourth media information in the next period after the initial period.

At the end of the contract advertising campaign:

Reward = -abs(actual number of views for the contracted ad – planned number of views for the contracted ad).

Here, abs represents the absolute value function. Contract advertising effectiveness refers to metrics such as CTR and CVR of contract ads. Alpha and beta range from 0 to 1, representing the simulator's weights, and are not limited here.

For example, suppose there is a contract ad A that has booked 1 million views for today. By 6:00 PM, it has already been viewed 900,000 times. During the 6:00 PM to 6:05 PM time slot, it participated in the mixed ad placement 300,000 times, with 100,000 views and an average click-through rate (CTR) of 0.02. When this contract ad competes with other ads during the 6:00 PM to 6:05 PM time slot, the winning ad has an average eCPM of 10 yuan, even though it lost 200,000 views.

Therefore, in the time slice from 18:00 to 18:05, Reward = 10w + alpha * 10w * 0.02 + beta * 20w * 10.

Then, during the 18:05-18:10 time slot, the contracted ad received 100,000 views. Other than that, the ad's performance remained the same as during the 18:00-18:05 time slot, meaning the number of failed CTR and PK ads and the average ECPM of the winning ad remained unchanged.

Therefore, in the time slice from 18:05 to 18:10, Reward = -10w + alpha * 10w * 0.02 + beta * 20w * 10.

In general, the main idea behind model training is as follows: Two models are used, one for predicting Actions (such as the media information evaluation model described in the text), denoted as Model A, and the other for predicting Rewards (such as the effect evaluation model and critic model described in the text), denoted as Model B. The main process is as follows: First, Model A uses the State obtained from the previous time slice to predict the Action for the current time slice. Then, Model A, based on the current State and the predicted Action, feeds the current State and the predicted Action into Model B to obtain the predicted Reward. Finally, the current State and the actual Action are fed into Model B to obtain the actual Reward.

The optimization objective of Model B is to make the predicted reward as close as possible to the actual reward. The optimization objective of Model A is to output an Action, feed it into Model B, and obtain a predicted reward that is as close to the actual reward as possible. During training, Model B's predictions become increasingly accurate, which is equivalent to Model A's optimization objective being to obtain a true reward that is as close as possible to the actual reward after outputting an Action.

The overall model training process is as follows:

First, obtain the training sample set. The offline training process for the game-round-based RL hybrid ranking model is as follows:

Step 1: For each contracted advertisement, obtain information about the contracted advertisement, including the number of views (such as the target number of views for third-party media information), the period of the contract, etc.

Step 2: Divide the predetermined period into time slices (e.g., 5-minute intervals).

Step 3, for each time slice, perform the following steps:

Step a: Collect the initial State for the current time slice.

Step b involves inputting the collected State into the RL model to obtain the predicted Action. This predicted Action corresponds to the actual playback evaluation value of the next time period after the initial time period described in the text.

Step c: Input the predicted Action and the mixed log of the contract advertisement in the current time slice into the simulator to obtain the Reward, the end marker "done", and the subsequent state - the termination State.

Step d: Save <Initial State, Action, Reward, Terminating State> as a sample (that is, store the initial state, predicted action, reward and termination state of each time slice as a sample). If the number of samples reaches a certain threshold, then perform RL model training (the RL model training process is described below).

In step e, if the end marker "done" is true, return to step 1.

If the end marker "done" is false in step f, then return to step a.

Step 4: If all contracts have been trained, the process ends.

The initial state of the current time slice is the initial state of the current time slice, which is the state at the end of the previous time slice. The ending state of the time slice is the state corresponding to the end of the current time slice (i.e., the initial state of the next time slice).

It should be noted that the current time slice can correspond to the next time segment after the initial time segment described in the text. In <Initial State, Action, Reward, Terminating State>, the Initial State corresponds to the sample target data of the initial time segment described in the text; the Action corresponds to the actual playback evaluation value of the next time segment after the initial time segment described in the text; the Reward corresponds to the actual evaluation performance representation value of the next time segment after the initial time segment described in the text; and the Terminating State corresponds to the sample target data of the first time segment of the initial time segment described in the text, where the first time segment is the time segment after the next time segment.

Optionally, when storing samples, each contract advertisement can use the State of the previous time slice, the Action predicted in the current time slice, the Reward obtained from the simulator in the current time slice, and the State of the next time slice (i.e., the <Initial State, Action, Reward, Terminating State> described in the text) as a sample. All samples stored in this way are denoted as sample A, which is the initially stored sample.

It should be noted that when training the RL model, all data in sample A are used as real data.

Steps a-c above are a process of generating multiple samples based on real historical data of the contract advertisement. The purpose is to enrich the sample. All generated samples are used as real samples for training the RL model. Through steps a-c, samples of the contract advertisement in each time slice can be generated. When the contract advertisement finishes playing and the generated samples exceed a certain threshold, the generated samples can be used to train the RL model.

Specifically, the RL model training process is as follows (taking the classic Action-critic configuration as an example):

Step 1: Train the critic model:

Step a, obtain the sample <Initial State, Action, Reward, Terminating State>;

Step b: Input the termination State into the Action model to obtain Action_Next;

Step c: Input <Termination State, Action_Next> into the critic model to obtain Reward_Next;

Step d: Calculate Final_Reward1 = Reward + lambda * Reward_Next;

Step e: Input <Initial State, Action> into the critc model to obtain Final_Reward2;

Step f: Train the critic model with input <initial State, Action> and loss = the difference between Final_Reward1 and Final_Reward2;

In training the critic model, the value of lambda is in the range of 0-1, and the decay weight of the long-term reward (i.e., reward_Next) is not limited here.

Step 2: Train the Action model:

Step a: Input the <Initial State> into the Action model to obtain Action_New;

Step b: Input <Initial State, Action_New> into the critc model to obtain Final_Reward3;

Step c: Train the Action model with the input <Initial State> and the loss -Final_Reward3.

The long-term rewards (i.e., Reward_Next, Final_Reward2, and Final_Reward3) are obtained corresponding to the rewards for all time slices.

It should be noted that the critic model is used to predict rewards (i.e., the effect evaluation model), while the action model is used to predict actions (i.e., the media information evaluation model).

In this context, `InitialState` in `<InitialState, Action_New>` corresponds to the target sample data of the initial time period described in the text; `Action_Next` corresponds to the predicted playback evaluation value of the next time period described in the text; and `Final_Reward3` corresponds to the first prediction evaluation performance representation value corresponding to the third media information described in the text. The sum of all `-Final_Reward3` values is the first total training loss.

In the `<Termination State, Action_Next>` section, `Action_Next` corresponds to the second predicted playback evaluation value of the third media information described in the text during the first time period; `Reward_Next` corresponds to the second predicted evaluation effect representation value of the third media information described in the text; `Final_Reward1` corresponds to the first evaluation effect representation value described in the text; and `Final_Reward2` corresponds to the second evaluation effect representation value described in the text. The difference between `Final_Reward1` and `Final_Reward2` is the difference between the first and second evaluation effect representation values.

When training the critic model, the goal is to minimize the difference between the loss Final_Reward1 and Final_Reward2, thereby increasing the accuracy of the critic model's reward predictions. Similarly, when training the action model, the goal is to minimize the loss -Final_Reward3 (i.e., to make Final_Reward3 closer to positive infinity), thereby improving the action model's action prediction performance.

According to one aspect of the present invention, a media information processing method is provided. Optionally, the media information processing method may be applied, but is not limited to, the application environment shown in FIG2. As shown in FIG2, the media information processing method may be applied, but is not limited to, the media information processing system in the environment shown in FIG2, wherein the media information processing system may include, but is not limited to, a user terminal 101, a network 102, and a server 103. The user terminal 101 can communicate with the server 103 through the network 102, and the user terminal 101 can initiate an Internet request to the server 103 through the network. An Internet application runs on the user terminal 101, which can provide Internet services to the user, such as providing services such as watching videos, browsing web pages, searching for information, and playing games. On the Internet application, the user can watch interstitial media information, such as advertisements. The Internet application can be a web application, an application program (APP), etc. The user terminal 101 includes a human-computer interaction screen 1011, a processor 1012, and a memory 1013. The human-computer interaction screen 1011 is used for the user to initiate Internet requests and can also be used for the user to watch media information to be played. Processor 1012 is used to handle operations related to the user's internet request. Memory 1013 is used to store data related to the internet request. Server 103 includes database 1031 and processing engine 1032.

As shown in Figure 2, the specific process of determining the media information to be played in the media information processing method of this application is as follows: steps S1-S9:

S1, the user initiates an Internet request through the Internet application in the user terminal 101 and sends the Internet request to the server 103 through the network 102.

S2, after receiving an internet request, server 103 responds to the internet request, such as returning the video corresponding to the internet request to user terminal 101. Simultaneously, the processing engine 1032 in server 103 obtains the target data of each first media information in the candidate media information set for the previous time period. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, playback competition information of the media information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information. The database 1031 in server 103 stores the target data.

S3, for each of the aforementioned first media information, the processing engine 1032 in the server 103 determines the first playback evaluation value of the aforementioned first media information in the current time period based on the target data of the aforementioned first media information in the previous time period.

S4, the processing engine 1032 of server 103 obtains the second playback evaluation value of each of the above-mentioned second media information in the current time period.

S5, the processing engine 1032 in server 103 determines the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the first media information and the second playback evaluation value corresponding to each of the second media information. The media information to be played is then sent to user terminal 101 via network 102.

S6, when the user terminal 101 receives the media information to be played, it plays the media information to be played on the human-computer interaction screen 1011 of the user terminal 101.

It is understood that the above is only one example, and this embodiment is not limited here.

The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server or server cluster providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The network can include, but is not limited to, wired networks and wireless networks. Wired networks include local area networks (LANs), metropolitan area networks (MANs), and wide area networks (WANs). Wireless networks include Bluetooth, Wi-Fi, and other networks that enable wireless communication. User terminals can be smartphones (such as Android phones, iOS phones, etc.), tablets, laptops, digital radio receivers, MIDs (Mobile Internet Devices), PDAs (Personal Digital Assistants), desktop computers, in-vehicle terminals (such as in-vehicle navigation terminals), smart speakers, smartwatches, etc. User terminals and servers can be directly or indirectly connected via wired or wireless communication, but are not limited to these methods. Specific details can be determined based on the actual application scenario requirements and are not limited here.

Referring to Figure 3, which is a flowchart illustrating a media information processing method provided in an embodiment of this application, the method can be executed by any electronic device, such as a server or a user terminal, or it can be completed through interaction between the user terminal and the server. Optionally, it can be executed by the server. As shown in Figure 3, the media information processing method provided in this embodiment of the application includes the following steps:

S301. Obtain the target data of each first media information in the candidate media information set in the previous time period. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of the historical playback volume in the previous time period, playback competition information of media information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information.

S302, for each of the aforementioned first media information, based on the target data of the aforementioned first media information in the previous time period, determine the first playback evaluation value of the aforementioned first media information in the current time period.

S303, obtain the second playback evaluation value of each of the above-mentioned second media information in the current time period, wherein, for the playback evaluation value corresponding to the above-mentioned first media information or second media information, the playback evaluation value represents the probability of the media information being promoted;

S304, based on the first playback evaluation value corresponding to each of the aforementioned first media information and the second playback evaluation value corresponding to each of the aforementioned second media information, determine the media information to be played in the current time period from the aforementioned candidate media information set.

Optionally, the candidate media information set includes at least one first type of first media information (such as contract advertising) and at least one second type of second media information (such as performance advertising).

For each target data corresponding to the first media information, the target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of the historical playback volume of the previous time period, media information playback competition information, or target playback volume.

The traffic prediction information refers to the traffic prediction information for the current time period and subsequent time periods corresponding to this first media information. This first media information can be traffic prediction information for different locations, such as the traffic prediction information for users in first-tier cities, etc., which is not limited here.

The historical play count mentioned above can refer to the historical play count of the first media information in the previous time period, such as the first media information being played 100,000 times in the previous time period. The target play count is the pre-set play count of the first media information, such as 1 million.

In one optional embodiment, the aforementioned playback competition information includes at least one of the following:

The historical click-through rate of each of the aforementioned primary media information within the previous time period;

The historical click-through rate of each of the aforementioned secondary media information within the previous time period;

The conversion rate of each of the aforementioned primary media messages in the previous time period;

The conversion rate of each of the aforementioned secondary media messages in the previous time period;

The exposure rate of each of the aforementioned primary media outlets in the previous time period;

The exposure rate of each of the aforementioned secondary media messages in the previous time period;

The playback evaluation value of each of the aforementioned primary media information in the previous time period;

The playback evaluation value of each of the aforementioned secondary media information in the previous time period.

Optionally, the exposure rate is used to characterize the win rate of the first media message, and is determined by the exposure of the first media message in the previous time period. The playback evaluation value is the bid of the first media message in the previous time period. The bid can also be understood as scoring the contracted advertisement to obtain the quality score of the contracted advertisement.

Based on the acquired target data, the first playback evaluation value of the first media information within the current time period can be determined. Then, the second playback evaluation values of each second media information participating in this mixed ranking competition within the current time period are acquired. According to the first and second playback evaluation values, the media information to be played in the current time period is determined from the candidate media set and played.

For example, let's take playing contract ads in the ad slot at the beginning of a popular TV series as an example. The goal is to achieve the reserved number of views for each of the 10 contract ads in that ad slot within one day (i.e., 00:00-24:00). For example, the reserved number of views for contract ad 1 is 1 million, the reserved number of views for contract ad 2 is 2 million, and so on, and the reserved number of views for contract ad 10 is 1.5 million.

Taking Contract Ad 1 as an example, the specific steps are as follows:

Step 1: Divide the day's delivery cycle into fine-grained time slices, such as 5 minutes.

Step 2: During the evening time slot, fewer people watch the popular drama, resulting in less traffic. Therefore, you can set a lower bid or even 0.

Step 3: During the daytime time slot, such as 9:00-9:05, obtain the play count of the contract ad 1 in the previous time slot (e.g., 50,000) (i.e., the historical play count of the previous time slot), the competition situation in the mixed playback stage (i.e., the competition situation of the other 9 contract ads besides contract ad 1, and the competition situation of other performance ads participating in the ad playback competition in this time slot) (i.e., the above media information playback competition information), and the corresponding traffic prediction information of the contract ad (e.g., more traffic from 12:30-13:00, more traffic from 20:00-22:30, etc.). Assuming that based on this information, Ad 1 has a high win rate between 9:00 and 9:05, has 850,000 remaining plays, and will generate significant traffic after the 9:00-9:05 time slot, then the bid for the 9:00-9:05 time slot can remain the same as the bid for the 8:55-9:00 time slot, and at least one performance ad that can generate substantial revenue should be introduced. Alternatively, compared to the 8:55-9:00 time slot, the bid for the 9:00-9:05 time slot can be increased, and one or more performance ads that can generate substantial revenue should be introduced.

Once the bid for Contract Ad 1 is determined, Contract Ad 1 will compete with 9 other Contract Ads and 1 or more other performance Ads during the 9:00-9:05 time slot. During the competition, in addition to the bid, factors such as the duration of the competing ads, the type of ads, the maximum number/duration limit of ads that can be played in the ad slot, and the same advertiser filtering logic (i.e., ads from the same advertiser cannot be played in the same ad slot during the same time slot) will be considered. The successful bidders will be ranked and played in the ad slot at the beginning of the aforementioned popular drama in the order of their ranking.

Step 4: Following the above method, bid, sort, and play the ads for each time slot until the predetermined number of plays is reached or the campaign period for that day is completed.

As shown in Figure 4a, assuming the media information to be played according to the above method consists of successfully bid contract ads 1, 3, 4, and 5, and performance ads 2 and 5, these successfully bid media information will be played in rotation at the beginning of the popular drama. In the target display interface shown in Figure 4a, the total duration of the ads 1, 3, 4, 5, 2, and 5 is 120 seconds. Currently playing in this target display interface is information about performance ad 5 (i.e., "Car A, your best choice!" shown in the figure). After performance ad 5 finishes playing, contract ads 1, 3, 4, 5, and 2 can be played.

As shown in Figure 4b, assuming the media information to be played according to the above method consists of successfully bid contract ads 1, 3, 4, and 5, and performance ads 2 and 5, these successfully bid media information will be played in rotation on the advertising slots at the beginning of the popular drama. In the target display interface shown in Figure 4b, information about contract ads 1, 4, and 2 is currently being played. After contract ads 1, 4, and 2 are played, information about contract ads 3, 5, and 5 can be played.

In other words, one or more advertisements can be played simultaneously on the target playback interface, without limitation. The examples shown in Figures 4a and 4b illustrate advertisements within a video; in practical applications, the specific type of advertisement is not limited.

In this embodiment, target data for each first media information in the candidate media information set in the previous time period is obtained. This target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, media information competition information, or target playback volume. For each first media information, a first playback evaluation value for the current time period can be determined based on the obtained target data from the previous time period. Then, combined with the obtained second playback evaluation values for each second media information in the current time period, the first and second media information in the candidate media information set can be sorted to determine the media information to be played in the current time period. Using this method, the first playback evaluation value for each first media information can be determined based on the target data, and the media information to be played in the current time period can be determined based on the first and second playback evaluation values. This effectively utilizes target data to determine the playback evaluation value of the first media information and uses the playback evaluation value to determine the media information to be played, improving the rationality of displaying the media information to be played.

In an optional embodiment, the playback competition information includes the playback evaluation value of the previous time period. For any of the first media information, determining the first playback evaluation value of the first media information in the current time period based on the target data of the first media information includes:

Based on the aforementioned target data, the playback evaluation value of the aforementioned first media information in the previous time period is adjusted to obtain the first playback evaluation value of the aforementioned first media information in the current time period.

Optionally, after obtaining the target data, the historical playback evaluation value of the first media information in the previous time period can be adjusted based on the target data to obtain the first playback evaluation value of the first media information in the current time period.

Specifically, the first playback evaluation value of the first media information in the current time period can be obtained by comprehensively considering the bidding situation, exposure rate, click rate, and conversion rate of the first media information in the previous period, the playback evaluation value of other second media information, the playback evaluation value of other first media information, the traffic prediction information of the first media information, the number of unfinished playbacks of the first media information, the ecpm distribution of the second media information on the corresponding traffic, and the distribution of the number of competitors of other first media information.

Taking advertising as an example, factors such as the competitive environment during the mixed-broadcast phase can be comprehensively considered, including the number of contract ads participating in the mixed-broadcast, the bid distribution of performance ads, the bid distribution of other contract ads, win rate, CTR/CVR distribution, etc. Traffic prediction factors include the number of unfinished plays for the contract ad, the ECPM distribution of performance ads on the corresponding traffic, and the distribution of competing contract ads. The delivery status factors include the pre-ordered volume of contract ads and the number of ads already exposed.

For example, if the target number of views for the first media message has not yet been reached in the current time period, but the first media message performed well in the previous time period, such as having a relatively high click-through rate and/or conversion rate, thus meeting the set requirements, then the historical view evaluation value of the previous time period can be kept unchanged, or the historical view evaluation value can be slightly reduced to obtain the first view evaluation value for the current time period.

If the target number of views for the first media message in the current time period has not been reached, and the performance of the first media message in the previous time period was relatively poor, such as a relatively low click-through rate and/or conversion rate, and the set requirements were not met, then the historical playback evaluation value of the previous time period can be increased to obtain the first playback evaluation value for the current time period.

If the target number of views for the first media message has not yet been reached in the current time period, but the first media message performed well in the previous time period, such as having a relatively high click-through rate and/or conversion rate, meeting the set requirements, and traffic prediction information indicates that there will be significant traffic generation after the current time period, then the historical playback evaluation value of the previous time period can be kept unchanged, or the historical playback evaluation value can be slightly reduced to obtain the first playback evaluation value for the current time period.

If the target number of views for the first media message has not been reached in the current time period, the performance of the first media message in the previous time period was relatively poor, such as a relatively low click-through rate and/or conversion rate, and the set requirements have not been met, and the traffic prediction information indicates that there will be no significant traffic generation after the current time period, then the historical playback evaluation value can be increased based on the historical playback evaluation value of the previous time period to obtain the first playback evaluation value for the current time period.

The playback evaluation value of other media information besides the first media information can also be considered. If the playback evaluation value of other media information is high (i.e., the bid is high), it indicates that the competition is relatively fierce. If the number of unfinished playbacks of the first media information is not much, the playback evaluation value of the current period can be reduced based on the historical playback evaluation value of the previous period. If the number of unfinished playbacks of the first media information is still relatively large, the playback evaluation value of the current period can be increased based on the historical playback evaluation value of the previous period.

It is understood that the above are only a few possible examples. In practical applications, various factors that may affect the number of plays and the eCPM value will be comprehensively considered to adjust the play evaluation value (i.e., bid) for the current time period. This embodiment does not impose any limitations here.

In one possible embodiment, the above method further includes:

Determine the probability of each of the aforementioned primary media information being played in the current time period;

The above-mentioned determination of the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the aforementioned first media information and the second playback evaluation value corresponding to each of the aforementioned second media information includes:

Based on the first playback evaluation value corresponding to each of the aforementioned first media information, the aforementioned probability of being played, and the second playback evaluation value corresponding to each of the aforementioned second media information, the media information to be played for the current time period is determined from the aforementioned candidate media information.

Optionally, in practical applications, when determining the media information to be played, in addition to considering the playback evaluation value (i.e., the bid), it is also necessary to consider the probability of the media information being played. That is, by comprehensively considering the first playback evaluation value and the probability of being played for each first media information, as well as the second playback evaluation value for each second media information, the media information to be played for the current time period is determined from the candidate media information.

In an optional embodiment, determining the probability of each of the first media information being played in the current time period includes:

For each of the aforementioned first media information, obtain the display evaluation data corresponding to the current time period for the aforementioned first media information, wherein the aforementioned display evaluation data refers to information that affects the probability of the aforementioned first media information being played;

For each of the aforementioned first media information, the probability of the aforementioned first media information being played in the current time period is determined based on the display evaluation data corresponding to the aforementioned first media information.

Optionally, the evaluation data can be relevant information that can affect the probability of the first media information being played. The probability of the first media information being played in the current time period can be determined by displaying the evaluation data.

In one optional embodiment, for any of the aforementioned first media information, the aforementioned display evaluation data includes at least one of the following:

Profile data of the target user corresponding to the current terminal device;

The device information of the current terminal device;

The attribute information of the aforementioned primary media information;

The current time information corresponding to the aforementioned terminal devices;

The current location information corresponding to the aforementioned terminal devices;

The above target users correspond to the behavioral statistics of the current terminal device.

Optionally, the target user profile data corresponding to the current terminal device can be understood as the user's age, gender, education level, interests, etc. Device-related information for the current terminal device can be understood as the network status of the current terminal device, the type of user device, such as 5G phones, 4G phones, 3G phones, 2G phones, and the corresponding product brands for each type of phone. The evaluation data displayed includes the current time information corresponding to the terminal device (e.g., 3 PM, 10 AM), and the current location information corresponding to the terminal device (e.g., Beijing, Shanghai, Qinghai-Tibet, etc.). The behavioral statistics of the target user corresponding to the current terminal device can be understood as the user's behavioral statistical characteristics (e.g., the number of videos watched, the number of advertisements viewed, and the time spent using the terminal device that day).

By displaying evaluation information, the probability of the primary media information being played can be determined.

The embodiments of this application can comprehensively consider factors such as playback evaluation value and probability of being played to determine the media information to be played, thereby improving the accuracy of determining the media information to be played.

In an optional embodiment, the determination of the first playback evaluation value of each of the first media information in the current time period based on the target data of the first media information in the previous time period is achieved through a media information evaluation model, which is trained in the following manner:

Obtain a training sample set, wherein each training sample in the training sample set includes a sample target data of a first type of third media information in the initial time period;

The sample target data of each of the aforementioned third media information in the initial time period are input into the initial information evaluation model to obtain the predicted playback evaluation value of each of the aforementioned third media information in the next time period in the initial time period.

For each of the aforementioned third media information, based on the sample target data of the aforementioned third media information in the initial time period and the predicted playback evaluation value in the next time period, the first predicted evaluation effect characterization value corresponding to the aforementioned third media information is determined.

Based on the first prediction effect evaluation characterization value corresponding to each of the above-mentioned third media information, the first total training loss corresponding to the above information evaluation model is determined.

The information evaluation model is repeatedly trained based on the aforementioned training samples and the first total training loss until the total training loss meets the preset first training termination condition, thus obtaining the aforementioned media information evaluation model.

Optionally, the predicted playback evaluation value is obtained through an initial information evaluation model, and the predicted playback evaluation value represents the probability of the aforementioned third media information being promoted.

The first playback evaluation value of the first media information is determined through a media information evaluation model, which is trained in the following way:

Taking advertising as an example, real historical data, such as the previous month's detailed scheduling logs (i.e., records of the relevant historical playback data of the advertisement), can be used as the training sample set. Alternatively, as described above, the sample set stored in the form of sample A can also be used as the training sample set; no restrictions are imposed here.

For the specific training process of this media information evaluation model, please refer to the specific process of training the Action model mentioned above, which will not be repeated here.

In this embodiment of the application, the bidding action is determined by the state information (State) through this reinforcement learning method. Then, the bidding action is evaluated by the reward mechanism (Reward) to determine whether the bid is reasonable. This allows the model to learn continuously until it is able to make reasonable bids, indicating that the model training is complete. The model automatically bids to determine the media information to be played, which effectively solves the problem of maximizing the volume of contract ads and the ecpm of performance ads when performing mixed ranking of ads. This achieves the optimization of the mixed ranking system and improves the rationality of mixed ranking.

In an optional embodiment, each of the above-mentioned training samples further includes the actual playback evaluation value of the third media information in the next time period after the initial time period, the actual evaluation effect representation value in the next time period after the initial time period, and the sample target data of the first time period of the initial time period, wherein the first time period is the next time period after the next time period.

Specifically, for each of the aforementioned third media information, the first predicted evaluation effect representation value corresponding to the aforementioned third media information is determined based on the sample target data and predicted playback evaluation value of the aforementioned third media information in the initial time period. This is achieved through an effect evaluation model, which is trained in the following manner:

The sample target data of each of the aforementioned third media information in the first time period are input into the initial information evaluation model to obtain the second predicted playback evaluation value of each of the aforementioned third media information in the first time period;

The sample target data of each of the above-mentioned third media information in the first period of the initial period and the corresponding second predicted playback evaluation value are input into the initial effect evaluation model to obtain the second predicted evaluation effect characterization value corresponding to each of the above-mentioned third media information.

For each of the aforementioned third media information, based on the aforementioned actual evaluation effect characterization value and the aforementioned second predicted evaluation effect characterization value, the first evaluation effect characterization value corresponding to the aforementioned third media information is determined. Based on the sample target data of the aforementioned third media information in the initial period and the actual playback evaluation value of the next period in the initial period, the second evaluation effect characterization value corresponding to the aforementioned third media information is obtained through the effect evaluation model.

Based on the first and second evaluation effect representation values corresponding to each of the above-mentioned third media information, the second total training loss corresponding to the above-mentioned effect evaluation model is determined.

Based on the aforementioned training samples and the aforementioned second total training loss, the performance evaluation model is repeatedly trained until the preset second training termination condition is met.

Optionally, the specific training process for this performance evaluation model can be found in the previous section on training the critic model, and will not be repeated here.

In an optional embodiment, each of the above-mentioned training samples further includes a playback effect evaluation parameter for at least one second-type fourth media information corresponding to the above-mentioned third media information in the next time period after the above-mentioned initial time period. For any of the above-mentioned third media information, the above-mentioned true evaluation effect representation value is obtained in the following manner:

Obtain the target play count of the aforementioned third-party media information and the play count of the next time period corresponding to the initial time period;

Based on the target playback volume and the playback volume, determine the playback effect evaluation parameters of the third media information.

Based on the playback effect evaluation parameters of the aforementioned third media information and the playback effect evaluation parameters of each of the corresponding fourth media information, the aforementioned true evaluation effect representation value is determined.

Optionally, for the specific calculation method of the above-mentioned true evaluation effect characterization value, please refer to the calculation method of the reward in the text, which will not be repeated here.

The electronic devices (including but not limited to user terminals and terminal devices) involved in the embodiments of this application can be nodes in a blockchain system, and the servers involved (which can be blockchain servers) can also be stored in blockchain nodes. The media information processing system involved in the embodiments of this application can be a distributed system formed by connecting clients and multiple blockchain nodes (any form of computing device in the network, such as servers and user terminals) through network communication.

Taking a distributed system as an example of a blockchain system, refer to Figure 5. Figure 5 is a schematic diagram of the structure of an optional distributed system 100 applied to a blockchain system according to an embodiment of the present invention. It consists of multiple nodes 200 (any form of computing device in the network, such as servers or user terminals) and clients 300. The nodes form a peer-to-peer (P2P) network. The P2P protocol is an application layer protocol running on top of the Transmission Control Protocol (TCP). In the distributed system, any machine, such as a server or terminal, can join and become a node. A node includes a hardware layer, a middleware layer, an operating system layer, and an application layer.

Refer to Figure 5 for the functions of each node in the blockchain system, which include:

1) Routing: A basic function of nodes used to support communication between nodes.

In addition to routing capabilities, nodes can also have the following functions:

2) Applications are deployed in the blockchain to implement specific business needs. They record data related to the implementation of functions to form record data, carry digital signatures in the record data to indicate the source of the task data, and send the record data to other nodes in the blockchain system. When other nodes successfully verify the source and integrity of the record data, they add the record data to a temporary block.

For example, the business logic implemented by the application includes:

2.1) A wallet is used to provide the function of conducting electronic currency transactions, including initiating transactions (i.e., sending the transaction record of the current transaction to other nodes in the blockchain system; after other nodes successfully verify the transaction, they store the transaction record data in the temporary block of the blockchain as a response to acknowledge the validity of the transaction; of course, the wallet also supports querying the remaining electronic currency in the electronic currency address;

2.2) Shared ledger, used to provide functions such as storage, query and modification of ledger data. It sends the record data of the operation on the ledger data to other nodes in the blockchain system. After the other nodes verify the validity, as a response to acknowledge the validity of the ledger data, they store the record data in a temporary block. They can also send confirmation to the node that initiated the operation.

2.3) Smart contracts are computerized protocols that can execute the terms of a contract. They are implemented through code deployed on a shared ledger that executes when certain conditions are met. Based on actual business needs, the code is used to complete automated transactions, such as querying the logistics status of goods purchased by a buyer and transferring the buyer's electronic money to the merchant's address after the buyer signs for the goods. Of course, smart contracts are not limited to executing contracts for transactions; they can also execute contracts for processing received information.

3) A blockchain consists of a series of blocks that are sequentially generated. Once a new block is added to the blockchain, it will not be removed. The blocks contain the data submitted by the nodes in the blockchain system.

Referring to Figure 6, which is a schematic diagram of an optional block structure provided by an embodiment of the present invention, each block includes the hash value of the transaction records stored in this block (the hash value of this block) and the hash value of the previous block. The blocks are connected through their hash values to form a blockchain. Additionally, the block may also include information such as a timestamp when it was generated. A blockchain is essentially a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains relevant information used to verify the validity of the information (anti-counterfeiting) and to generate the next block.

Referring to Figure 7, Figure 7 is a schematic diagram of the structure of a media information processing device provided in an embodiment of this application. The media information processing device 1 provided in this embodiment of the application includes:

The target data acquisition module 11 is used to acquire the target data of each first media information in the candidate media information set in the previous time period. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, playback competition information of media information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information.

The playback evaluation value processing module 12 is used to determine the first playback evaluation value of the first media information in the current time period based on the target data of the first media information in the previous time period for each of the first media information.

The playback evaluation value processing module 12 is used to obtain the second playback evaluation value of each of the above-mentioned second media information in the current time period, wherein, for the playback evaluation value corresponding to the above-mentioned first media information or second media information, the playback evaluation value represents the probability of the media information being promoted;

The media information processing module 13 is used to determine the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the first media information and the second playback evaluation value corresponding to each of the second media information.

The aforementioned playback competition information includes at least one of the following pieces of information from each of the candidate media information sets:

Click-through rate in the previous time period;

Conversion rate in the previous period;

Exposure rate in the previous period;

The playback evaluation value of the previous time period.

In an optional embodiment, the playback evaluation value processing module described above is used for:

Based on the aforementioned target data, the historical playback evaluation value of the aforementioned first media information in the previous time period is adjusted to obtain the first playback evaluation value of the aforementioned first media information in the current time period.

In an optional embodiment, the above-mentioned media information processing module is further configured to:

Determine the probability of each of the aforementioned primary media information being played in the current time period;

The above-mentioned determination of the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the aforementioned first media information and the second playback evaluation value corresponding to each of the aforementioned second media information includes:

Based on the first playback evaluation value corresponding to each of the aforementioned first media information, the aforementioned probability of being played, and the second playback evaluation value corresponding to each of the aforementioned second media information, the media information to be played for the current time period is determined from the aforementioned candidate media information.

In an optional embodiment, the above-mentioned media information processing module is further configured to:

For each of the aforementioned first media information, obtain the display evaluation data corresponding to the current time period for the aforementioned first media information, wherein the aforementioned display evaluation data refers to information that affects the probability of the aforementioned first media information being played;

For each of the aforementioned first media information, the probability of the aforementioned first media information being played in the current time period is determined based on the display evaluation data corresponding to the aforementioned first media information.

In one optional embodiment, for any of the aforementioned first media information, the aforementioned display evaluation data includes at least one of the following:

Profile data of the target user corresponding to the current terminal device;

The device information of the current terminal device;

The attribute information of the aforementioned primary media information;

The current time information corresponding to the aforementioned terminal devices;

The current location information corresponding to the aforementioned terminal devices;

The above target users correspond to the behavioral statistics of the current terminal device.

In an optional embodiment, for each of the aforementioned first media information, determining the first playback evaluation value of the first media information in the current time period based on the target data of the first media information in the previous time period is achieved through a media information evaluation model. The device further includes a training module, which is used to train the media information evaluation model. The media information evaluation model is trained in the following manner:

Obtain a training sample set, wherein each training sample in the training sample set includes a sample target data of a first type of third media information in the initial time period;

The sample target data of each of the aforementioned third media information in the initial time period are input into the initial information evaluation model to obtain the predicted playback evaluation value of each of the aforementioned third media information in the next time period in the initial time period.

For each of the aforementioned third media information, based on the sample target data of the aforementioned third media information in the initial time period and the predicted playback evaluation value in the next time period, the first predicted evaluation effect characterization value corresponding to the aforementioned third media information is determined.

Based on the first prediction effect evaluation characterization value corresponding to each of the above-mentioned third media information, the first total training loss corresponding to the above information evaluation model is determined.

The information evaluation model is repeatedly trained based on the aforementioned training samples and the first total training loss until the total training loss meets the preset first training termination condition, thus obtaining the aforementioned media information evaluation model.

In an optional embodiment, each training sample further includes the actual playback evaluation value of the third media information in the next time period after the initial time period, the actual evaluation effect representation value in the next time period after the initial time period, and the sample target data of the first time period of the initial time period, wherein the first time period is the time period after the next time period. For each piece of third media information, based on the sample target data and predicted playback evaluation value of the third media information in the initial time period, the first predicted evaluation effect representation value corresponding to the third media information is determined through an effect evaluation model. The effect evaluation model is trained by the training module, which is used to:

The sample target data of each of the aforementioned third media information in the first time period are input into the initial information evaluation model to obtain the second predicted playback evaluation value of each of the aforementioned third media information in the first time period;

The sample target data of each of the above-mentioned third media information in the first period of the initial period and the corresponding second predicted playback evaluation value are input into the initial effect evaluation model to obtain the second predicted evaluation effect characterization value corresponding to each of the above-mentioned third media information.

For each of the aforementioned third media information, based on the aforementioned actual evaluation effect characterization value and the aforementioned second predicted evaluation effect characterization value, the first evaluation effect characterization value corresponding to the aforementioned third media information is determined. Based on the sample target data of the aforementioned third media information in the initial period and the actual playback evaluation value of the next period in the initial period, the second evaluation effect characterization value corresponding to the aforementioned third media information is obtained through the effect evaluation model.

Based on the first and second evaluation effect representation values corresponding to each of the above-mentioned third media information, the second total training loss corresponding to the above-mentioned effect evaluation model is determined.

Based on the aforementioned training samples and the aforementioned second total training loss, the performance evaluation model is repeatedly trained until the preset second training termination condition is met.

In an optional embodiment, each of the above-mentioned training samples further includes a playback effect evaluation parameter for at least one second-type fourth media information corresponding to the above-mentioned third media information in the next time period after the above-mentioned initial time period. For any of the above-mentioned third media information, the above-mentioned true evaluation effect representation value is obtained by the above-mentioned training module in the following manner:

Obtain the target play count of the aforementioned third-party media information and the play count of the next time period corresponding to the initial time period;

Based on the target playback volume and the playback volume, determine the playback effect evaluation parameters of the third media information.

Based on the playback effect evaluation parameters of the aforementioned third media information and the playback effect evaluation parameters of each of the corresponding fourth media information, the aforementioned true evaluation effect representation value is determined.

In practice, the media information processing device 1 can execute the implementation methods provided by each step in Figure 3 above through its built-in functional modules. For details, please refer to the implementation methods provided by each step above, which will not be repeated here.

In this embodiment, target data for each first media information in the candidate media information set in the previous time period is obtained. This target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, media information competition information, or target playback volume. For each first media information, a first playback evaluation value for the current time period can be determined based on the obtained target data from the previous time period. Then, combined with the obtained second playback evaluation values for each second media information in the current time period, the first and second media information in the candidate media information set can be sorted to determine the media information to be played in the current time period. Using this method, the first playback evaluation value for each first media information can be determined based on the target data, and the media information to be played in the current time period can be determined based on the first and second playback evaluation values. This effectively utilizes target data to determine the playback evaluation value of the first media information and uses the playback evaluation value to determine the media information to be played, improving the rationality of displaying the media information to be played.

The above mainly describes that the execution subject is hardware to implement the media information processing method in this application. However, the execution subject of the media information processing method in this application is not limited to hardware. The execution subject of the media information processing method in this application can also be software. The aforementioned media information processing device can be a computer program (including program code) running on a computer device. For example, the media information processing device is an application software. The device can be used to execute the corresponding steps in the method provided in the embodiments of this application.

In some embodiments, the media information processing apparatus provided in this invention can be implemented using a combination of hardware and software. As an example, the media information processing apparatus provided in this invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the media information processing method provided in this invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

In other embodiments, the media information processing device provided in the embodiments of the present invention can be implemented in software. The media information processing device 1 shown in FIG8 can be software in the form of programs and plug-ins, and includes a series of modules, including a target data acquisition module 11, a playback evaluation value processing module 12, and a media information processing module to be played 13; wherein, the target data acquisition module 11, the playback evaluation value processing module 12, and the media information processing module to be played are used to implement the media information processing method provided in the embodiments of the present invention.

Referring to Figure 8, which is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, the electronic device 1000 in this embodiment may include a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the electronic device 1000 may also include a user interface 1003 and at least one communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1004 may be a high-speed RAM or a non-volatile memory, such as at least one disk storage device. The memory 1005 may optionally be at least one storage device located remotely from the processor 1001. As shown in Figure 8, the memory 1005, as a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application.

In the electronic device 1000 shown in Figure 8, the network interface 1004 provides network communication functions; the user interface 1003 is mainly used to provide an input interface for users; and the processor 1001 can be used to call computer programs stored in the memory 1005.

It should be understood that in some feasible implementations, the processor 1001 described above may be a central processing unit (CPU), which may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

In practice, the aforementioned electronic device 1000 can execute the implementation methods provided in the steps shown in Figure 3 above through its built-in functional modules. For details, please refer to the implementation methods provided in the steps above, which will not be repeated here.

This application also provides a computer-readable storage medium storing a computer program that is executed by a processor to implement the methods provided in the various steps of FIG3. For details, please refer to the implementation methods provided in the above steps, which will not be repeated here.

The aforementioned computer-readable storage medium can be an internal storage unit of the task processing apparatus provided in any of the foregoing embodiments, such as a hard disk or memory of an electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device. The aforementioned computer-readable storage medium can also include magnetic disks, optical disks, read-only memory (ROM), or random access memory (RAM), etc. Furthermore, the computer-readable storage medium can include both internal storage units and external storage devices of the electronic device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in any of the possible embodiments shown in FIG3.

The terms "first," "second," etc., used in the claims, description, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that comprises a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses. References to "embodiment" herein mean that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The presentation of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. The term "and/or" as used in the description and appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Those skilled in the art can implement the described functions using different methods for each specific application, but such implementations should not be considered beyond the scope of this application.

The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims (12)

1. A media information processing method, characterized in that it includes: Obtain the target data of each first media information in the candidate media information set in the previous time period. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, playback competition information of media information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information. For each piece of the first media information, based on the target data of the first media information in the previous time period, determine the first playback evaluation value of the first media information in the current time period; Obtain the second playback evaluation value of each of the second media information in the current time period, wherein, for the playback evaluation value corresponding to the first media information or the second media information, the playback evaluation value represents the probability of the media information being promoted; Based on the first playback evaluation value corresponding to each of the first media information and the second playback evaluation value corresponding to each of the second media information, the media information to be played in the current time period is determined from the candidate media information set. 2. The method according to claim 1, wherein the playback competition information includes at least one of the following information from each media information in the candidate media information set: Click-through rate in the previous time period; Conversion rate in the previous period; Exposure rate in the previous period; The playback evaluation value of the previous time period. 3. The method according to claim 2, wherein the playback competition information includes the playback evaluation value of the previous time period, and for any first media information, determining the first playback evaluation value of the first media information in the current time period based on the target data of the first media information includes: Based on the target data, the playback evaluation value of the first media information in the previous time period is adjusted to obtain the first playback evaluation value of the first media information in the current time period. 4. The method according to claim 1, characterized in that the method further comprises: Determine the probability of each of the first media information being played in the current time period; The step of determining the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the first media information and the second playback evaluation value corresponding to each of the second media information includes: Based on the first playback evaluation value corresponding to each of the first media information, the probability of being played, and the second playback evaluation value corresponding to each of the second media information, the media information to be played in the current time period is determined from the candidate media information. 5. The method according to claim 4, wherein determining the probability of each of the first media information being played in the current time period includes: For each piece of the first media information, obtain the display evaluation data corresponding to the current time period for the first media information, wherein the display evaluation data refers to information that affects the probability of the first media information being played; For each piece of the first media information, the probability of the first media information being played in the current time period is determined based on the display evaluation data corresponding to the first media information. 6. The method according to claim 5, wherein for any of the first media information, the display evaluation data includes at least one of the following: Profile data of the target user corresponding to the current terminal device; The device information of the current terminal device; The attribute information of the first media information; The current time information corresponding to the terminal device; The current location information corresponding to the terminal device; The target user corresponds to the behavioral statistics of the current terminal device. 7. The method according to any one of claims 1 to 6, characterized in that, for each piece of first media information, determining the first playback evaluation value of the first media information in the current time period based on the target data of the first media information in the previous time period is achieved through a media information evaluation model, wherein the media information evaluation model is trained in the following manner: Obtain a training sample set, wherein each training sample in the training sample set includes sample target data of a first type of third media information in the initial time period; The sample target data of each of the third media information in the initial time period are input into the initial information evaluation model to obtain the predicted playback evaluation value of each of the third media information in the next time period in the initial time period. For each of the third media information, based on the sample target data of the third media information in the initial time period and the predicted playback evaluation value in the next time period, the first predicted evaluation effect characterization value corresponding to the third media information is determined; Based on the first prediction effect evaluation characterization value corresponding to each of the third media information, the first total training loss corresponding to the initial information evaluation model is determined. The initial information evaluation model is repeatedly trained based on each training sample and the first total training loss until the total training loss meets the preset first training termination condition, thereby obtaining the media information evaluation model. 8. The method according to claim 7, wherein each training sample further includes the actual playback evaluation value of the third media information in the next time period after the initial time period, the actual evaluation effect representation value in the next time period after the initial time period, and the sample target data of the first time period, wherein the first time period is the next time period after the next time period of the initial time period; Specifically, for each piece of third media information, the first predicted evaluation effect representation value corresponding to the third media information is determined based on the sample target data of the third media information in the initial time period and the predicted playback evaluation value in the next time period. This is achieved through an effect evaluation model, which is trained in the following way: The sample target data of each of the third media information in the first time period are input into the initial information evaluation model to obtain the second predicted playback evaluation value of each of the third media information in the first time period; The sample target data of each of the third media information in the first time period and the corresponding second predicted playback evaluation value are input into the initial effect evaluation model to obtain the second predicted evaluation effect characterization value corresponding to each of the third media information. For each of the third media information, based on the actual evaluation effect characterization value and the second predicted evaluation effect characterization value, a first evaluation effect characterization value corresponding to the third media information is determined. Based on the sample target data of the third media information in the initial period and the actual playback evaluation value of the next period in the initial period, a second evaluation effect characterization value corresponding to the third media information is obtained through the effect evaluation model. Based on the first and second evaluation effect characterization values corresponding to each of the third media information, the second total training loss corresponding to the effect evaluation model is determined. Based on each training sample and the second total training loss, the performance evaluation model is repeatedly trained until the preset second training termination condition is met. 9. The method according to claim 8, wherein each training sample further includes a playback effect evaluation parameter for at least one second-type fourth media information corresponding to the third media information in the next time period after the initial time period, and for any third media information, the true evaluation effect representation value is obtained in the following manner: Obtain the target playback count of the third media information and the playback count of the next time period corresponding to the initial time period; Based on the target number of plays and the number of plays, determine the playback effect evaluation parameters of the third media information; The true evaluation effect representation value is determined based on the playback effect evaluation parameters of the third media information and the playback effect parameters of each fourth media information corresponding to the third media information. 10. A media information processing apparatus, characterized in that the apparatus comprises: The target data acquisition module is used to acquire the target data of each first media information in the candidate media information set in the previous time period. The target data includes at least one of traffic prediction information and playback-related data. The playback-related data includes at least one of historical playback volume in the previous time period, playback competition information of media information, or target playback volume. The candidate media information set includes at least one first type of first media information and at least one second type of second media information. The playback evaluation value processing module is used to determine the first playback evaluation value of the first media information in the current time period based on the target data of the first media information in the previous time period for each first media information. The playback evaluation value processing module is used to obtain the second playback evaluation value of each of the second media information in the current time period, wherein, for the playback evaluation value corresponding to the first media information or the second media information, the playback evaluation value represents the probability of the media information being promoted; The media information processing module is used to determine the media information to be played in the current time period from the candidate media information set based on the first playback evaluation value corresponding to each of the first media information and the second playback evaluation value corresponding to each of the second media information. 11. An electronic device, characterized in that it comprises a processor and a memory, the processor and the memory being interconnected; The memory is used to store computer programs; The processor is configured to perform the method as described in any one of claims 1 to 9 when the computer program is invoked. 12. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, the computer program being executed by a processor to implement the method according to any one of claims 1 to 9.