CN119831704A - Product intelligent recommendation system based on cloud platform - Google Patents
Product intelligent recommendation system based on cloud platform Download PDFInfo
- Publication number
- CN119831704A CN119831704A CN202411948754.XA CN202411948754A CN119831704A CN 119831704 A CN119831704 A CN 119831704A CN 202411948754 A CN202411948754 A CN 202411948754A CN 119831704 A CN119831704 A CN 119831704A
- Authority
- CN
- China
- Prior art keywords
- medical product
- target
- target user
- semantic
- product
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a product intelligent recommendation system based on a cloud platform, which is used for intelligently generating a recommendation probability value of a target candidate medical product based on semantic flow field alignment matching representation between semantic coding features of the target user portrait and semantic coding features of target candidate medical product materials by acquiring user portrait information of the target user object, acquiring material information of the target candidate medical product from a plurality of candidate medical products, and carrying out semantic coding on the user portrait of the target user object and the material information of the target candidate medical product by utilizing an artificial intelligence-based data analysis and processing mode, thereby determining whether the target candidate medical product is taken as the target medical product. In this way, consistency and objectivity of the recommended results can be ensured. Meanwhile, the latest behavior and preference change of the user can be captured in real time, the recommendation model is updated in time, and the recommendation result is ensured to always meet the current requirement of the user.
Description
Technical Field
The application relates to the field of intelligent recommendation, in particular to a product intelligent recommendation system based on a cloud platform.
Background
In recent years, with the rapid expansion of the medical beauty market and the increasing demand of consumers for personalized services, conventional medical product recommendation methods have failed to meet the expectations of users. Consumers not only want to obtain product recommendations meeting their unique needs, but also expect to find the most suitable medical solutions in a convenient and efficient way.
Patent CN111859122B proposes a method, apparatus, electronic device and readable storage medium for recommending medical products, which receives images or videos uploaded by users, analyzes and determines body part problems and generates demand information. Based on this information, matching candidate products are selected from a plurality of medical products. And combining the characteristics of consumption history, search, treatment record and the like of the user, and further screening out the most suitable target medical product for recommendation.
In the patent, weight values are set for candidate medical products according to characteristic information of a user, and target medical products are obtained based on the weight values. However, the setting of weight values is highly dependent on the experience and assumptions of the designer, and different experts may set different weights based on different backgrounds and understandings, resulting in deviations in the recommended results, which subjectivity makes it difficult for the system to form a unified standard, affecting the consistency and accuracy of the recommendation. In addition, the preferences and demands of the user are dynamically changed, and the weight value is usually set based on historical data, so that the latest demands of the user cannot be reflected in time. This means that even if the interests or conditions of the user change, the system may still recommend according to outdated data, resulting in recommendation results that are inconsistent with the current needs of the user.
Accordingly, an optimized intelligent recommendation of products is desired.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides a product intelligent recommendation system based on a cloud platform, which is used for intelligently generating a recommendation probability value of a target candidate medical product based on semantic field alignment matching representation between semantic coding features of a target user image and semantic coding features of target candidate medical product materials by acquiring user image information of the target user object, acquiring material information of the target candidate medical product from a plurality of candidate medical products, and carrying out semantic coding on the user image of the target user object and the material information of the target candidate medical product by utilizing an artificial intelligence-based data analysis and processing mode, thereby determining whether the target candidate medical product is taken as the target medical product. In this way, consistency and objectivity of the recommended results can be ensured. Meanwhile, the latest behavior and preference change of the user can be captured in real time, the recommendation model is updated in time, and the recommendation result is ensured to always meet the current requirement of the user.
According to one aspect of the application, an intelligent product recommendation system based on a cloud platform is provided, which comprises a user data acquisition module, a data transmission module, a demand information generation module, a medical product candidate module and a medical product recommendation module, wherein the user data acquisition module is used for acquiring data to be identified, the data to be identified is input by a target user object, the data transmission module is used for transmitting the data to be identified to a data processor based on the cloud platform, the demand information generation module is used for analyzing the data to be identified in the data processor based on the cloud platform to obtain demand information of the target user object, the medical product candidate module is used for determining a plurality of candidate medical products based on the demand information, and the medical product recommendation module is used for determining target medical products from the plurality of candidate medical products.
Compared with the prior art, the intelligent product recommendation system based on the cloud platform provided by the application has the advantages that through acquiring the user portrait information of the target user object and acquiring the material information of the target candidate medical product from the plurality of candidate medical products, semantic coding is carried out on the user portrait of the target user object and the material information of the target candidate medical product by utilizing an artificial intelligence-based data analysis and processing mode, so that the recommendation probability value of the target candidate medical product is intelligently generated based on the semantic flow field alignment matching representation between the semantic coding characteristics of the target user portrait and the semantic coding characteristics of the material of the target candidate medical product, and whether the target candidate medical product is used as the target medical product is determined. In this way, consistency and objectivity of the recommended results can be ensured. Meanwhile, the latest behavior and preference change of the user can be captured in real time, the recommendation model is updated in time, and the recommendation result is ensured to always meet the current requirement of the user.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.
FIG. 1 is a block diagram of a cloud platform-based product intelligent recommendation system in accordance with an embodiment of the present application;
fig. 2 is a schematic data flow diagram of a product intelligent recommendation system based on a cloud platform according to an embodiment of the present application;
fig. 3 is a block diagram of a traditional Chinese medicine product recommendation module of the intelligent product recommendation system based on the cloud platform according to an embodiment of the application;
Fig. 4 is a block diagram of a user-product feature matching unit in a cloud platform-based product intelligent recommendation system according to an embodiment of the present application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
As used in the specification and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
Although the present application makes various references to certain modules in a system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and/or server. The modules are merely illustrative, and different aspects of the systems and methods may use different modules.
A flowchart is used in the present application to describe the operations performed by a system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously, as desired. Also, other operations may be added to or removed from these processes.
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
In the technical scheme of the application, a product intelligent recommendation system based on a cloud platform is provided. Fig. 1 is a block diagram of a cloud platform-based product intelligent recommendation system according to an embodiment of the present application. Fig. 2 is a schematic data flow diagram of a product intelligent recommendation system based on a cloud platform according to an embodiment of the present application. As shown in fig. 1 and 2, the intelligent product recommendation system 300 based on the cloud platform according to the embodiment of the application comprises a user to-be-identified data acquisition module 310 for acquiring to-be-identified data input by a target user object, a to-be-identified data transmission module 320 for transmitting the to-be-identified data to a data processor based on the cloud platform, a demand information generation module 330 for analyzing the to-be-identified data at the data processor based on the cloud platform to obtain demand information of the target user object, a medical product candidate module 340 for determining a plurality of candidate medical products based on the demand information, and a medical product recommendation module 350 for determining a target medical product from the plurality of candidate medical products.
In particular, the user to-be-identified data obtaining module 310 is configured to obtain to-be-identified data input by the target user object. In an example of the present application, the system supports the user uploading a photo or video file that has been taken, also allowing the user to use the camera directly for real-time shooting or recording. After receiving the data uploaded by the user, the system can initially judge the validity of the data, so as to ensure that the data is suitable for further processing. In particular, the system will check if the data contains an image of the body part (e.g., face, head, etc.) available for analysis, and if the data is invalid, feedback an error message to the user and prompt for reentry. In addition, the quality of the image or video, such as sharpness, lighting conditions, etc., is also assessed to ensure the accuracy of the subsequent analysis.
Once the data is validated, the system automatically recognizes the body parts (such as face, legs, etc.) involved in the image or video by computer vision techniques, and simultaneously recognizes the basic information of the user such as age, sex, etc., so as to obtain the corresponding standard image as a reference for comparison. Based on the above identification, the system retrieves from the database standard images matching the determined body position, age and gender, and establishes a frame of reference so that the system can accurately identify problems (e.g., spots, skin conditions, etc.) present on the user's body. And finally, comparing the data to be identified with the standard image by the system, and generating the requirement information of the user according to the difference between the data to be identified and the standard image. This process may include specific facial feature extraction (e.g., wrinkles, acne, etc.), problem localization (specifically indicating a specific problem with a certain part of the user's body), and detailed feature acquisition (for certain specific problems, such as spot size, skin dryness, etc., the system may continue to analyze in depth to obtain more detailed user demand information).
In particular, the data to be identified transmitting module 320 is configured to transmit the data to be identified to a data processor based on a cloud platform. In an example of the present application, after a user submits an image or video through the system interface, the data to be identified is first temporarily stored by the local device. In order to ensure the security of data transmission, the system performs primary encryption processing on the received data, and adopts an industry standard encryption algorithm such as AES (advanced encryption standard) or RSA (Rivest-Shamir-Adleman) to prevent the data from being intercepted or tampered during the transmission process. The encrypted data is then packetized into a format suitable for network transport, such as a multimedia file stream under the HTTP/HTTPs protocol. Then, the system establishes a safe and reliable connection channel for sending the packaged data to the cloud server. This process typically relies on sophisticated cloud computing service platforms that provide rich API interfaces and service options that support a variety of programming languages and development environments so that developers can easily integrate data transfer functions. At the same time, they also have powerful network infrastructure and load balancing capabilities, ensuring stable data transmission performance even during network fluctuations or peak hours.
In particular, the requirement information generating module 330 is configured to analyze, at the cloud platform-based data processor, the data to be identified to obtain requirement information of the target user object. In the example of the present application, when an image or video submitted by a user is transmitted to a cloud server through a secure encrypted channel, the image or video is first sent to a preprocessing module. In this module, the system performs preliminary processing on the received data, including but not limited to format conversion, resizing, and quality optimization. These steps are intended to lay a good foundation for further analysis, ensuring consistency and availability of the input data. For example, for picture files of different resolutions or encoding formats, the system will be uniformly converted to standard specifications, and for video streams, it may be necessary to extract key frames in order to complete the analysis more quickly. Next, the analysis phase of the core is entered. The key here is how to efficiently identify and understand the body part features contained in an image or video and the differences between them and the ideal state. For this purpose, the system employs a series of advanced computer vision techniques and deep learning models to perform fine-tuning analysis. Specifically, an algorithm such as Convolutional Neural Network (CNN) is used to automatically locate a face or other relevant body region in the image, and to mark important anatomical landmark points, such as the positions of eyes, nose, and mouth. This step is crucial for a more accurate problem detection later, as it provides a basic frame of reference for the user's physical condition. Once the body region of interest is determined, the system will further perform problem detection and feature extraction. In this process, various methods and techniques may be combined, such as skin texture analysis based on statistical models, color distribution statistics, morphological transformation, etc., to identify specific skin problems (such as spots, acnes), wrinkles, or irregular contours. In addition, to improve accuracy, some auxiliary tools, such as 3D reconstruction techniques, can be introduced to generate a more stereoscopic and intuitive facial model, thereby better assessing some subtle changes. Meanwhile, the system also tries to identify the basic information of the user such as age, sex and the like, and prepares for the next step of standard image matching.
In particular, the medical product candidate module 340 is configured to determine a plurality of candidate medical products based on the demand information. That is, after the demand information of the user is generated, a plurality of candidate medical products are extracted from the medical product database based on the demand information. In an example of the present application, the system will first obtain material information for all available medical products, including but not limited to a list of the components of the product, skin fit, expected effect, method of use, user assessment, etc. These textual descriptions are then converted to a structured data format using Natural Language Processing (NLP) techniques to facilitate subsequent analysis. For example, through Word embedding models such as Word2Vec or BERT, component descriptions of products can be converted into high-dimensional vector representations, thereby capturing semantic similarity between different components. Meanwhile, for some quantitative attributes, such as applicable age ranges, price intervals and the like, numerical values are directly stored. Next, the system calculates a degree of matching between the user demand information and the material information of each medical product. Specifically, the system can find the product with corresponding efficacy according to specific requirement points (such as freckle removal, wrinkle resistance, moisture preservation and the like) of the user, and evaluate whether the components of the product are suitable for the skin type of the user. In addition, the safety and side effect risks of the product are considered, and the recommended product is ensured to be effective, safe and reliable. To improve the accuracy of the matching calculation, the system may introduce advanced algorithms, such as Support Vector Machines (SVMs), random Forest (Random Forest), or neural networks (Neural Network), to simulate expert-level decision logic.
In particular, the medical product recommendation module 350 is configured to determine a target medical product from the plurality of candidate medical products. In one specific example of the present application, as shown in fig. 3, the medical product recommendation module 350 includes a user portrait information obtaining unit 351 configured to obtain user portrait information of a target user object, a material information obtaining unit 352 configured to obtain material information of a target candidate medical product from the plurality of candidate medical products, a user product information encoding unit 353 configured to perform semantic encoding on the user portrait information of the target user object and the material information of the target candidate medical product to obtain a target user portrait semantic encoding feature and a target candidate medical product material semantic encoding feature, respectively, and a user-product feature matching unit 354 configured to perform semantic flow field guided feature response matching on the target user portrait semantic encoding feature and the target candidate medical product material semantic encoding feature to obtain a target user-medical product fine granularity alignment response matching characterization, and a product judging unit configured to determine whether to use the target candidate medical product as the target medical product 355 based on the target user-medical product fine granularity alignment response matching characterization.
Specifically, the user portrait information acquisition unit 351 is configured to acquire user portrait information of a target user object. In particular, the user portrayal information includes basic information, consumption information, search information, and treatment information. In one example, the underlying information is such as the user's age, gender, geographic location, and the like. These basic information helps the system to initially screen out medical product categories that may be suitable for the user. For example, users of different ages and sexes may be more concerned about different cosmetic problems, young people may be more concerned about acne and uneven skin tone, and middle aged people may be more concerned about anti-aging and wrinkle treatment. Therefore, by knowing the basic attributes of the user, the system can reduce the range of candidate products and improve the pertinence of recommendation, and the consumption information reflects the historical purchasing behavior and consumption habit of the user. Through analysis of the user's past purchase records, the system is able to identify the brand, price range, and specific component or efficacy type of product that the user prefers. For example, if a user often purchases a high-end brand of whitening product, the system may tend to select new products with the same price segment and similar functionality at the time of recommendation. The recommendation based on the historical behavior mode not only improves the acceptance of users, but also increases the probability of sales conversion, and in addition, the search information is one of important components of the user portrait. The user's search history can reveal their current points of interest and potential needs even though these needs have not been translated into actual purchasing behavior. For example, if users frequently search for information about spot removal, but have not purchased related products, this may indicate that they are strongly concerned about the problem, but still in the sightseeing stage. At this time, the system can excite the interests of the user and guide the user to make decisions by pushing some spot-removing products at proper time or well-evaluated. Treatment information is also non-negligible and encompasses medical services or treatment regimens that the user has received. This part of the information is particularly valuable for the system, as it is directly related to the physical condition and therapeutic effect of the user. For example, if the user had previously performed a laser spot removal procedure, the system would need to take this into account when recommending new spot care products and preferably select those products that do not conflict with the original treatment or can assist. In this way, not only the profession and science of recommendation are improved, but also a more comprehensive solution is provided for the user.
Specifically, the material information obtaining unit 352 is configured to obtain material information of a target candidate medical product from the plurality of candidate medical products. It will be appreciated that the acquisition of material information for candidate medical products can help the system to fully understand the characteristics of each product for finer screening. Such material information encompasses a wide range of content including, but not limited to, a list of ingredients of the product, skin application, intended effect, method of use, user evaluation, and the like. For example, for a type of freckle-removing cream, the system needs to know not only the main components (such as nicotinamide, tranexamic acid and the like) but also the action mechanism of the components and the adaptability of the components to different skin types. In this way, the system, when faced with users having specific needs, can choose the most suitable product according to its skin type and personal preferences. Further, the detailed material information helps the system evaluate the matching between each candidate product and the user's needs. By comparing the personalized needs of the user (such as spot removal, wrinkle resistance, moisture retention, etc.) with the functional characteristics of the products, the system can calculate the similarity or difference between the two, thereby determining which products are most likely to meet the user's expectations. For example, if the user's demand is for a whitening product for sensitive skin, the system will prioritize those products that are marked as mild formulations, no irritating ingredients, and give higher weight to their actual effect scores. The matching mode based on data driving not only improves the accuracy of recommendation, but also enhances the user experience. In addition, the material information can be obtained to provide more reference bases for users, so that the trust feeling of the users on the recommendation result is increased. When the system recommends a medical product to a user, the user can be more clearly informed of the background information, advantages and potential risks of the product by the aid of the detailed description data. For example, showing user ratings can help other consumers determine the true effect of the product, while providing detailed instructions can guide the user to apply correctly, avoiding adverse consequences due to incorrect operation. Therefore, not only is the recommendation effectiveness improved, but also better decision support is provided for the user.
Specifically, the user product information encoding unit 353 is configured to perform semantic encoding on the user portrait information of the target user object and the material information of the target candidate medical product, so as to obtain a target user portrait semantic encoding feature and a target candidate medical product material semantic encoding feature. In the technical scheme of the application, the user portrait information of the target user object passes through a user portrait semantic encoder comprising a Bert model to obtain a target user portrait semantic encoding feature vector as the target user portrait semantic encoding feature. Here, considering that the user portrait information of the target user object contains a great amount of rich semantic information and deep meaning, these information reflect the preference and requirement of the user. Therefore, in order to deeply mine language features and capture potential semantic meanings, in the technical scheme of the application, the user portrait information of the target user object is subjected to deep semantic understanding through a user portrait semantic encoder comprising a Bert model to extract key context semantic information and implicit semantic meanings so as to obtain target user portrait semantic encoding feature vectors. Similarly, considering that the material information of the target candidate medical product contains various aspects of contents, the information such as product description, component list, applicable crowd and the like of the target candidate medical product usually exist in a text form, and semantic dependency exists between contexts among the text information. Based on the above, in order to further capture and extract semantic information between the key contexts of the product, in the technical scheme of the application, the semantic encoder of the medical product containing the Bert model is used for carrying out semantic encoding on the material information of the target candidate medical product so as to extract richer semantic representation and obtain the semantic encoding feature vector of the material of the target candidate medical product, thus the feature vector of each medical product can accurately reflect the unique attribute of each medical product, and the recommendation correlation is improved.
Specifically, the user-product feature matching unit 354 is configured to perform feature response matching guided by a semantic flow field on the semantic coding feature of the target user portrait and the semantic coding feature of the target candidate medical product material to obtain a fine granularity alignment response matching characterization of the target user-medical product. Considering that the target user portrait semantic coding features and the target candidate medical product material semantic coding features respectively express the specific requirements of the user on medical products and the detailed semantic information of the products. Simply combining these features may result in loss of information or may not accurately reflect the actual situation. Based on the method, a semantic flow field guided feature response matching mechanism is introduced to perform response matching on the target user portrait semantic coding features and the target candidate medical product material semantic coding features, so that a target user-medical product fine granularity alignment response matching characterization is obtained. That is, by constructing a matching mechanism that captures the complex relationship between the user and the medical product, fine-grained matching from the user's needs to specific product characteristics can be achieved, including not only explicit needs (e.g., specific ingredients, applicable skin types) but also implicit preferences (e.g., treatment history or potential skin problems of the user), thereby more accurately identifying the user's unique needs and matching the most suitable product characteristics, thereby providing highly personalized recommendations, improving user satisfaction. In one embodiment of the present application, as shown in fig. 4, the user-product feature matching unit 354 includes a target user feature alignment subunit 3541 configured to perform semantic field-based feature alignment on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector to obtain an aligned target user portrait semantic coding feature vector and an aligned target candidate medical product material semantic coding feature vector, a target user portrait linear transformation subunit 3542 configured to perform linear transformation on the aligned target user portrait semantic coding feature vector to obtain a target user portrait query vector and a target user portrait value vector, a target candidate medical product material linear transformation subunit 3543 configured to perform linear transformation on the aligned target candidate medical product material semantic coding feature vector to obtain a target candidate medical product material key vector, and a target user feature fine grain matching subunit 3544 configured to perform scale response encoding on the target user portrait query vector, the target user portrait value and the target candidate medical product key vector to obtain a target user-response fine grain matching response characteristic medical product matching fine grain matching as the target user feature matching fine grain matching target.
More specifically, the target user product feature alignment subunit 3541 is configured to perform feature alignment based on a semantic field on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector to obtain an aligned target user portrait semantic coding feature vector and an aligned target candidate medical product material semantic coding feature vector. In the embodiment of the application, firstly, a target user-medical product semantic flow field between the target user image semantic coding feature vector and the target candidate medical product material semantic coding feature vector is constructed. The target user-medical product semantic stream field can capture a complex semantic relation space between the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector. By constructing the semantic flow field of the target user-medical product, the subsequent characteristic alignment process can more accurately identify and adjust the fine difference among the characteristics, thereby laying a foundation for deeper characteristic interaction. In a specific implementation, the process of constructing the semantic flow field between the first feature vector and the second feature vector generally comprises performing feature activation based on point convolution and Sigmoid function on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector to obtain a target user portrait semantic coding activation feature vector and a target candidate medical product material semantic coding activation feature vector, and performing semantic interaction based on convolution kernel attention on the target user portrait semantic coding activation feature vector and the target candidate medical product material semantic coding activation feature vector to obtain the target user-medical product semantic flow field. More specifically, a semantic flow field of the target user-medical product between the image semantic coding feature vector of the target user and the semantic coding feature vector of the material of the target candidate medical product is constructed according to the following semantic flow field construction formula, wherein the semantic flow field construction formula is as follows: Wherein, the method comprises the steps of, Is the target user portrayal semantic coding feature vector,Is the semantic coding feature vector of the target candidate medical product material,Is a point convolution code which is used to encode,Is the function of the activation and,Is the target user image semantic coding activation feature vector,Is the semantic coding activation feature vector of the target candidate medical product material,Is thatIs used to determine the transposed vector of (c),Is a matrix multiplication which is a function of the matrix,Is thatIs provided for the length of (a),Is a convolution code with a convolution kernel of 5x5,For a convolutional code with a convolutional kernel of 3 x3,Is an up-sampling operation and,Is a semantic flow field of a target user-medical product.
And then, carrying out feature alignment on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector based on the target user-medical product semantic flow field, so that target user portrait semantic coding features from different sources or modes and target candidate medical product material semantic coding features can be accurately matched in the same semantic space, and the aligned target user portrait semantic coding feature vector and the aligned target candidate medical product material semantic coding feature vector are obtained. The target user-medical product semantic flow field is used for carrying out feature alignment on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector according to the following feature alignment formula to obtain the aligned target user portrait semantic coding feature vector and the aligned target candidate medical product material semantic coding feature vector, wherein the feature alignment formula is as follows: Wherein, the method comprises the steps of, Is the aligned target user portrait semantic coding feature vector,Is the aligned semantic coding feature vector of the target candidate medical product material.
More specifically, the target user image linear transformation subunit 3542 and the target candidate medical product material linear transformation subunit 3543 are configured to perform linear transformation on the aligned target user image semantic coding feature vector to obtain a target user image query vector and a target user image value vector, and perform linear transformation on the aligned target candidate medical product material semantic coding feature vector to obtain a target candidate medical product material key vector. Wherein the query vector and the key vector are used to locate relevant features and the value vector represents specific information content. To ensure the validity of the transformation, the linear transformation matrix is usually defined by parameters learned during the training process, ensuring that the aligned feature vectors can maximize the preservation of the semantic information of the original features and have a stronger expressive power in the new feature space. More specifically, the aligned target user portrait semantic coding feature vector and the aligned target candidate medical product material semantic coding feature vector are linearly transformed by the following linear transformation formula to obtain the target user portrait query vector, the target user portrait value vector and the target candidate medical product material key vector, wherein the linear transformation formula is as follows: Wherein, the method comprises the steps of, AndA target user portrait query embedding matrix and a target user portrait query bias vector,Is the target user portrait query vector,AndThe target user image value embedding matrix and the target user image value offset vector,Is a vector of image values of the target user,AndThe target candidate medical product material key embedding matrix and the target candidate medical product material key offset vector are respectively,Is a target candidate medical product material key vector.
More specifically, the target user product feature fine granularity matching subunit 3544 is configured to perform fine granularity response encoding on the target user portrait query vector, the target user portrait value vector, and the target candidate medical product material key vector to obtain a target user-medical product fine granularity alignment response matching token vector as the target user-medical product fine granularity alignment response matching token vector. In one specific example of the present application, the target user portrait query vector, the target user portrait value vector, and the target candidate medical product material key vector are subjected to fine granularity response matching based on a converter structure to obtain the target user-medical product fine granularity alignment response matching characterization vector. The converter structure can process long-distance dependency relationship without depending on sequence order through a multi-head self-attention mechanism and a feedforward neural network, and model interaction among features. In the fine-grained response encoding process, the distribution of attention weights is guided by calculating the similarity between the query vector and the key vector, thereby determining which portions of the value vector should be emphasized or suppressed. And obtaining the fine granularity alignment response matching characterization vector of the target user-medical product after repeated overlapped attention mechanisms and nonlinear transformation. More specifically, the target user portrait query vector, the target user portrait value vector and the target candidate medical product material key vector are subjected to fine granularity response matching based on a converter structure in the following fine granularity response matching formula to obtain a fine granularity alignment response matching characterization vector of the target user-medical product, wherein the fine granularity response matching formula is as follows: Wherein, the method comprises the steps of, Is thatIs used to determine the transposed vector of (c),Is thatIs provided for the length of (a),Is thatThe function of the function is that,Is the matching characterization vector of the fine granularity alignment response of the target user-medical product.
Specifically, the product determining unit 355 is configured to determine whether to use the target candidate medical product as the target medical product based on the target user-medical product fine-granularity alignment response matching characterization. In the technical scheme of the application, a recommendation result is obtained based on the fine-granularity alignment response matching characterization vector of the target user-medical product, wherein the recommendation result is used for representing a recommendation probability value of the target candidate medical product, namely, the recommendation probability value of the target candidate medical product is intelligently generated by classifying the fine-granularity alignment response matching characterization of the target user-medical product obtained by performing semantic flow field response matching by using the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector. Therefore, the latest behavior and preference change of the user can be captured in real time, the recommendation model is updated in time, and the recommendation result is ensured to always meet the current requirement of the user. Further, the recommendation is compared to a predetermined threshold to determine whether to treat the target candidate medical product as a target medical product. Thus, by setting a reasonable threshold, it is ensured that only those medical products that are highly matched to the needs of the user will be ultimately recommended, which helps to improve the quality of the recommendation, ensuring that the recommended products do meet the needs and desires of the user.
In a preferred example, passing the target user-medical product fine-grained alignment response matching token vector through a classifier-based medical product recommender to obtain a recommendation result comprises:
Determining the number of non-zero eigenvalues in the target user-medical product fine-granularity alignment response matching characterization vector, and subtracting the number of non-zero eigenvalues from the length of the target user-medical product fine-granularity alignment response matching characterization vector to obtain a target user-medical product fine-granularity alignment response matching zero-dimensional numerical value ;
Calculating the target user-medical product fine-granularity alignment response matching zero-dimensional value minus one, and then matching zero-microscopic valueThe first target user-medical product fine-grain alignment response matching field fitting value obtained after being multiplied and divided by the target user-medical product fine-grain alignment response matching zero-dimensional numerical valueMatching field fit values with a second target user-medical product fine-grain alignment response;
Calculating the square root of the sum of squares of all eigenvalues of the target user-medical product fine-granularity alignment response matching characterization vector to obtain a target user-medical product fine-granularity alignment response matching pattern characterization value;
Matching zero-minim values with the target user-medical product fine-granularity alignment responseCalculating each eigenvalue of the target user-medical product fine-granularity alignment response matching characterization vector as an indexPower of (v) functionAnd multiplying the first target user-medical product fine-grain alignment response matching field fit valueTo obtain a target user-medical product fine granularity alignment response matching local representation value;
Calculating each eigenvalue of the target user-medical product fine-granularity alignment response matching characterization vectorMatching field fit values to the second target user-medical product fine-grain alignment responseAnd the target user-medical product fine granularity alignment response matching mode characterization valueTo obtain the target user-medical product fine-granularity alignment response matching center pointing value;
Matching the target user-medical product fine-granularity alignment response to a local representation valueMatching center-pointing values with the target user-medical product fine-grained alignment responseIs a weighted difference of (2)Matching the characteristic vector of the fine granularity alignment response of the composition optimized target user-medical product, and
And the optimized target user-medical product fine granularity alignment response matching characterization vector passes through a medical product recommender based on a classifier to obtain a recommendation result.
Here, under the condition that the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector respectively represent coding semantic features of target user portrait information and target candidate medical product material information, when feature interaction response matching based on semantic flow field modulation is performed, source sample semantic flow differential field modulation is not aligned, interaction response matching richness of the target user-medical product fine granularity alignment response matching characterization vector is caused, unstructured probability mapping is repeated, and accuracy of classification results is affected.
Therefore, based on a vector field fitting rule of a high-dimensional manifold, the isolated zero in the target user-medical product fine-granularity alignment response matching characterization vector is used as a micro dimension of the high-dimensional manifold so as to fix the normal direction pointing to the distribution center near the local coordinate represented by the characteristic value of the target user-medical product fine-granularity alignment response matching characterization vector, and therefore the effective alignment from the characteristic distribution mode of the target user-medical product fine-granularity alignment response matching characterization vector to the probability density micro field is realized to avoid invalid repetition in the probability mapping process of the semantic richness distribution of the target user-medical product fine-granularity alignment response matching characterization vector due to the unstructured characteristic of the semantic richness distribution of the target user-medical product fine-granularity alignment response matching characterization vector, and the accuracy of the recommendation result obtained by the medical product recommender based on the classifier is improved.
As described above, the cloud platform-based product intelligent recommendation system 300 according to the embodiment of the present application may be implemented in various wireless terminals, for example, a server or the like having a cloud platform-based product intelligent recommendation algorithm. In one possible implementation, the cloud platform-based product intelligent recommendation system 300 according to embodiments of the present application may be integrated into a wireless terminal as one software module and/or hardware module. For example, the intelligent cloud platform based product recommendation system 300 may be a software module in the operating system of the wireless terminal or may be an application developed for the wireless terminal, and of course, the intelligent cloud platform based product recommendation system 300 may also be one of a plurality of hardware modules of the wireless terminal.
Alternatively, in another example, the cloud platform based product intelligent recommendation system 300 and the wireless terminal may be separate devices, and the cloud platform based product intelligent recommendation system 300 may be connected to the wireless terminal through a wired and/or wireless network and transmit interaction information in a contracted data format.
The foregoing description of the embodiments of the present disclosure has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (9)
1. The intelligent product recommendation system based on the cloud platform is characterized by comprising a user to-be-identified data acquisition module, a to-be-identified data transmission module, a demand information generation module, a medical product candidate module and a medical product recommendation module, wherein the user to-be-identified data acquisition module is used for acquiring to-be-identified data input by a target user object, the to-be-identified data transmission module is used for transmitting the to-be-identified data to a data processor based on the cloud platform, the demand information generation module is used for analyzing the to-be-identified data in the data processor based on the cloud platform to obtain demand information of the target user object, the medical product candidate module is used for determining a plurality of candidate medical products based on the demand information, the medical product recommendation module is used for determining a target medical product from the plurality of candidate medical products, and the medical product recommendation module comprises:
a user portrait information acquisition unit for acquiring user portrait information of a target user object;
the material information acquisition unit is used for acquiring material information of target candidate medical products from the plurality of candidate medical products;
the user product information coding unit is used for respectively carrying out semantic coding on the user portrait information of the target user object and the material information of the target candidate medical product so as to obtain target user portrait semantic coding characteristics and target candidate medical product material semantic coding characteristics;
The user-product feature matching unit is used for carrying out feature response matching guided by semantic flow fields on the target user portrait semantic coding features and the target candidate medical product material semantic coding features so as to obtain target user-medical product fine granularity alignment response matching characterization;
and the product judging unit is used for determining whether the target candidate medical product is used as the target medical product based on the target user-medical product fine granularity alignment response matching characterization.
2. The cloud platform based product intelligent recommendation system of claim 1, wherein the user portrayal information comprises basic information, consumption information, search information, and treatment information.
3. The cloud platform based intelligent recommendation system for products of claim 2, wherein the user product information encoding unit comprises:
A target user portrait semantic coding subunit, configured to pass the user portrait information of the target user object through a user portrait semantic coder that includes a Bert model to obtain a target user portrait semantic coding feature vector as the target user portrait semantic coding feature;
The target candidate medical product material semantic coding subunit is used for enabling the material information of the target candidate medical product to pass through a medical product semantic coder containing a Bert model to obtain a target candidate medical product material semantic coding feature vector as the target candidate medical product material semantic coding feature.
4. The cloud platform based intelligent recommendation system for products of claim 3, wherein said user-product feature matching unit comprises:
the target user product feature alignment subunit is used for carrying out feature alignment based on a semantic field on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector so as to obtain an aligned target user portrait semantic coding feature vector and an aligned target candidate medical product material semantic coding feature vector;
The target user portrait linear transformation subunit is used for carrying out linear transformation on the aligned target user portrait semantic coding feature vector so as to obtain a target user portrait query vector and a target user portrait value vector;
the target candidate medical product material linear transformation subunit is used for carrying out linear transformation on the aligned target candidate medical product material semantic coding feature vector so as to obtain a target candidate medical product material key vector;
And the target user product feature fine granularity matching sub-unit is used for carrying out fine granularity response coding on the target user portrait query vector, the target user portrait value vector and the target candidate medical product material key vector so as to obtain a target user-medical product fine granularity alignment response matching representation vector as the target user-medical product fine granularity alignment response matching representation.
5. The cloud platform based intelligent recommendation system of claim 4, wherein said target user product feature alignment subunit comprises:
A second-level subunit is constructed by a target user product semantic flow field and is used for constructing a target user-medical product semantic flow field between the target user image semantic coding feature vector and the target candidate medical product material semantic coding feature vector;
And the feature alignment secondary subunit is used for carrying out feature alignment on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector based on the target user-medical product semantic flow field so as to obtain the aligned target user portrait semantic coding feature vector and the aligned target candidate medical product material semantic coding feature vector.
6. The cloud platform based product intelligent recommendation system of claim 5, wherein the target user product semantic flow field constructs a secondary subunit for:
Performing feature activation based on point convolution and Sigmoid function on the target user portrait semantic coding feature vector and the target candidate medical product material semantic coding feature vector to obtain a target user portrait semantic coding activation feature vector and a target candidate medical product material semantic coding activation feature vector;
and carrying out semantic interaction based on convolution kernel attention on the target user image semantic coding activation feature vector and the target candidate medical product material semantic coding activation feature vector to obtain the target user-medical product semantic flow field.
7. The intelligent product recommendation system based on the cloud platform of claim 6, wherein the target user product feature fine-granularity matching subunit is configured to perform fine-granularity response matching based on a converter structure on the target user portrait query vector, the target user portrait value vector, and the target candidate medical product material key vector to obtain the target user-medical product fine-granularity alignment response matching characterization vector.
8. The cloud platform based intelligent recommendation system for products of claim 7, wherein the product determination unit comprises:
The recommendation probability value generation subunit is used for obtaining a recommendation result based on the target user-medical product fine granularity alignment response matching characterization vector, wherein the recommendation result is used for representing the recommendation probability value of the target candidate medical product;
and the threshold comparison subunit is used for comparing the recommended result with a preset threshold to determine whether the target candidate medical product is taken as a target medical product.
9. The intelligent cloud platform based product recommendation system of claim 8, wherein said recommendation probability value generation subunit is configured to pass said target user-medical product fine granularity alignment response matching token vector through a classifier-based medical product recommender to obtain said recommendation result, said recommendation result being used to represent a recommendation probability value for said target candidate medical product.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411948754.XA CN119831704A (en) | 2024-12-27 | 2024-12-27 | Product intelligent recommendation system based on cloud platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411948754.XA CN119831704A (en) | 2024-12-27 | 2024-12-27 | Product intelligent recommendation system based on cloud platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN119831704A true CN119831704A (en) | 2025-04-15 |
Family
ID=95307555
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411948754.XA Withdrawn CN119831704A (en) | 2024-12-27 | 2024-12-27 | Product intelligent recommendation system based on cloud platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119831704A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119416898A (en) * | 2025-01-08 | 2025-02-11 | 吉林大学 | Intelligent auxiliary system and method for teaching gastrointestinal nursing |
| CN120370807A (en) * | 2025-04-17 | 2025-07-25 | 浙江瑞远智能机床股份有限公司 | Electrical control system of numerical control machine tool |
| CN120931332A (en) * | 2025-10-13 | 2025-11-11 | 上海七圣网络科技有限公司 | User portrait generation method and system for medical users |
-
2024
- 2024-12-27 CN CN202411948754.XA patent/CN119831704A/en not_active Withdrawn
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119416898A (en) * | 2025-01-08 | 2025-02-11 | 吉林大学 | Intelligent auxiliary system and method for teaching gastrointestinal nursing |
| CN120370807A (en) * | 2025-04-17 | 2025-07-25 | 浙江瑞远智能机床股份有限公司 | Electrical control system of numerical control machine tool |
| CN120931332A (en) * | 2025-10-13 | 2025-11-11 | 上海七圣网络科技有限公司 | User portrait generation method and system for medical users |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US12271814B2 (en) | Automatic digital content captioning using spatial relationships method and apparatus | |
| US11409791B2 (en) | Joint heterogeneous language-vision embeddings for video tagging and search | |
| CN119831704A (en) | Product intelligent recommendation system based on cloud platform | |
| US11699095B2 (en) | Cross-domain recommender systems using domain separation networks and autoencoders | |
| US11803872B2 (en) | Creating meta-descriptors of marketing messages to facilitate in delivery performance analysis, delivery performance prediction and offer selection | |
| CN115331150A (en) | Image recognition method, image recognition device, electronic equipment and storage medium | |
| WO2020020137A1 (en) | Commodity recommendation method, apparatus and system, and computer readable storage medium | |
| EP3285222A1 (en) | Facilitating television based interaction with social networking tools | |
| US20250061491A1 (en) | Chatbot with dynamic product advertisements | |
| WO2021155691A1 (en) | User portrait generating method and apparatus, storage medium, and device | |
| US12423368B2 (en) | Recommendation method and apparatus, training method and apparatus, device, and recommendation system | |
| Ma et al. | Landmark‐Based Facial Feature Construction and Action Unit Intensity Prediction | |
| CN112084954A (en) | Video target detection method and device, electronic equipment and storage medium | |
| CN116127080A (en) | Method for extracting attribute value of description object and related equipment | |
| CN115130493A (en) | Face deformation recommendation method, device, equipment and medium based on image recognition | |
| CN115129975A (en) | Recommendation model training method, recommendation device, recommendation equipment and storage medium | |
| US20220092361A1 (en) | System Improvement For Deep Neural Networks | |
| CN119336977A (en) | Method, device, storage medium and electronic device for constructing customer portrait | |
| KR102465453B1 (en) | A virtual makeup composition A.I. processing apparatus and a method using it | |
| CN116956183A (en) | Multimedia resource recommendation method, model training method, device and storage medium | |
| CN118051629A (en) | Content generation method, device, computer equipment and storage medium | |
| CN114926210A (en) | Comment generation method, device, equipment and computer-readable storage medium | |
| KR102744935B1 (en) | A method for providing style transfer images based on user information | |
| CN113297520A (en) | Page design auxiliary processing method and device and electronic equipment | |
| CN118210975A (en) | Object list recommendation method and device, electronic equipment and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Country or region after: China Address after: 310000 Hangzhou City, Binjiang District, Xixing Street, Jinxiu Linglong Mansion Building 1, Room 1113-2 (self-declared) Applicant after: Paz (Hangzhou) Technology Co.,Ltd. Address before: 310000 Zhejiang Province, Hangzhou City, Binjiang District, Changhe Street, Binwen Road 200, Building 1, 1st Floor, Room 1015 (self-declared) Applicant before: Dayou Xianxian (Hangzhou) Technology Co.,Ltd. Country or region before: China |
|
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20250415 |