CN111047044B

CN111047044B - ETA model training method, ETA estimation method, device and equipment

Info

Publication number: CN111047044B
Application number: CN201811185448.XA
Authority: CN
Inventors: 刘雨亭; 赵红超; 孟繁荣
Original assignee: Tencent Technology Shenzhen Co Ltd; Tencent Dadi Tongtu Beijing Technology Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd; Tencent Dadi Tongtu Beijing Technology Co Ltd
Priority date: 2018-10-11
Filing date: 2018-10-11
Publication date: 2022-02-01
Anticipated expiration: 2038-10-11
Also published as: CN111047044A

Abstract

The embodiment of the application discloses a training method of an ETA model, an ETA estimation method, a device and equipment. The method comprises the following steps: acquiring at least one piece of track data of a target user, wherein the at least one piece of track data corresponds to a road with at least one attribute; acquiring the speed ratio of the target user on the road with the at least one attribute according to the at least one piece of track data; and training an ETA personalized model for estimating the ETA of the target user according to the speed ratio characteristics of the target user. On one hand, the speed bit characteristics can be obtained by automatically mining the track data of the user by means of data analysis and processing, so that the defects of long time consumption and difficulty in ensuring accuracy existing in manual mining are overcome; on the other hand, the speed ratio characteristics can quantitatively represent the personalized behavior habits of the target user on the roads with different attributes, so that the ETA personalized model obtained through final training is more accurate when ETA estimation is performed on the target user.

Description

ETA model training method, ETA estimation method, device and equipment

Technical Field

The embodiment of the application relates to the technical field of machine learning, in particular to a training method of an ETA (Estimated Time of Arrival) model, an ETA estimation method, an ETA estimation device and ETA estimation equipment.

Background

The ETA estimation refers to the estimation of the time when the user reaches a specified destination. For example, in a navigation process, the time from the departure place to the destination of the user is predicted.

At present, the ETA estimation scheme based on machine learning in the industry mostly adopts an artificial feature engineering method, an ETA model is obtained by modeling through a regression model, and the ETA of a user is estimated through the ETA model. Conventional ETA models only consider route characteristics such as those including physical attributes of the route (such as attributes of the route's length, width, road grade, etc.), historical mining speed of the route, real-time speed of the route, etc. The traditional scheme only considers the route characteristics and does not consider the influence of the user on the ETA estimation result, so that the ETA estimation result is not accurate enough.

In the related art, on the basis of the route characteristics, user characteristics are further mined, and an ETA model is trained by fusing the two characteristics, so that an ETA personalized model for pre-estimating the ETA of a specific user can be obtained. The user characteristics may include: the user's own attributes (such as the user's age, gender, etc.), the user's familiarity with the route, the user's range of major active cities, etc.

However, the user features related to the related art are basically obtained by manual mining, the time consumption of the manual mining is long, and the accuracy of the feature values of the manual mining is difficult to guarantee, so that the estimation accuracy of the ETA personalized model obtained by final training is influenced.

Disclosure of Invention

The embodiment of the application provides a training method of an ETA model, an ETA estimation method, a device and equipment, so that the time consumption of feature value mining can be saved, and the estimation accuracy of the ETA personalized model can be improved. The technical scheme is as follows:

in one aspect, an embodiment of the present application provides a method for training an ETA model, where the method includes:

acquiring at least one piece of track data of a target user, wherein the at least one piece of track data corresponds to a road with at least one attribute;

acquiring the speed ratio of the target user on the road with the at least one attribute according to the at least one piece of track data; wherein the speed ratio is the ratio of the moving speed of the target user to the road passing speed;

training an ETA personalized model for estimating the ETA of the target user according to the speed ratio characteristics of the target user; wherein the speed bit characteristic of the target user comprises a speed ratio of the target user on the road of the at least one attribute.

On the other hand, an embodiment of the present application provides an ETA estimation method, which includes:

the method comprises the steps of obtaining travel information of a target user, wherein the travel information comprises a starting point position and an end point position;

determining a planned route from the start location to the end location;

obtaining route characteristics corresponding to the planned route and user characteristics of the target user, wherein the user characteristics of the target user comprise speed ratio characteristics of the target user, the speed ratio characteristics of the target user comprise speed ratios of the target user on roads with specified quantity of attributes, and the speed ratio is a ratio of the moving speed of the target user to the road running speed;

and calling the ETA personalized model of the target user, and obtaining an ETA estimation result corresponding to the planned route through the ETA personalized model according to the route characteristics corresponding to the planned route and the user characteristics of the target user.

In another aspect, an embodiment of the present application provides an ETA model training apparatus, where the apparatus includes:

the data acquisition module is used for acquiring at least one piece of track data of a target user, wherein the at least one piece of track data corresponds to a road with at least one attribute;

a speed ratio obtaining module, configured to obtain, according to the at least one piece of trajectory data, a speed ratio of the target user on the road with the at least one attribute; wherein the speed ratio is the ratio of the moving speed of the target user to the road passing speed;

the model training module is used for training an ETA personalized model for estimating the ETA of the target user according to the speed ratio characteristics of the target user; wherein the speed bit characteristic of the target user comprises a speed ratio of the target user on the road of the at least one attribute.

In another aspect, an embodiment of the present application provides an ETA estimation apparatus, where the apparatus includes:

the system comprises a journey acquisition module, a journey processing module and a journey processing module, wherein the journey acquisition module is used for acquiring journey information of a target user, and the journey information comprises a starting point position and an end point position;

a route planning module to determine a planned route from the start location to the end location;

a characteristic obtaining module, configured to obtain a route characteristic corresponding to the planned route and a user characteristic of the target user, where the user characteristic of the target user includes a speed ratio characteristic of the target user, the speed characteristic of the target user includes a speed ratio of the target user on a road with a specified number of attributes, and the speed ratio is a ratio of a moving speed of the target user to a road passing speed;

and the ETA estimation module is used for calling the ETA personalized model of the target user and obtaining the ETA estimation result corresponding to the planned route according to the route characteristics corresponding to the planned route and the user characteristics of the target user through the ETA personalized model.

In yet another aspect, an embodiment of the present application provides a computer device, which includes a processor and a memory, where the memory stores at least one instruction, at least one program, a set of codes, or a set of instructions, and the at least one instruction, the at least one program, the set of codes, or the set of instructions is loaded and executed by the processor to implement the ETA model training method or the ETA prediction method according to the above aspect.

In yet another aspect, an embodiment of the present application provides a computer-readable storage medium, in which at least one instruction, at least one program, a code set, or a set of instructions is stored, and the at least one instruction, the at least one program, the code set, or the set of instructions is loaded and executed by a processor to implement the ETA model training method or the ETA estimation method according to the above aspect.

In yet another aspect, the present application provides a computer program product, when being executed, for performing the ETA model training method or the ETA estimation method according to the above aspect.

The beneficial effects brought by the technical scheme provided by the embodiment of the application can include:

obtaining speed ratio characteristics of a target user by obtaining the speed ratio of the target user on various roads with different attributes, and then training an ETA personalized model for pre-estimating the ETA of the target user according to the speed ratio characteristics of the target user; on one hand, the speed bit characteristics can be obtained by automatically mining the track data of the user by means of data analysis and processing, so that the defects of long time consumption and difficulty in ensuring accuracy existing in manual mining are overcome; on the other hand, the speed ratio characteristics can reflect the difference between the moving speed of the target user on the roads with different attributes and the moving speed of most users, namely the personalized behavior habits of the target user on the roads with different attributes are quantitatively represented, so that the ETA personalized model obtained by final training is more accurate when ETA estimation is carried out on the target user.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a flow chart of a method for training an ETA model provided in one embodiment of the present application;

FIG. 2 illustrates a schematic diagram of several different road attributes;

FIG. 3 illustrates a schematic diagram of a probability distribution of speed ratios on a road of a certain attribute;

FIG. 4 illustrates a schematic diagram of a section division;

FIG. 5 illustrates a schematic diagram of a behavior distribution feature;

FIG. 6 illustrates a distribution diagram of topic description features of a user;

FIG. 7 is a schematic diagram illustrating behavior distribution characteristics corresponding to several subjects respectively;

FIG. 8 is a diagram illustrating an example of a user clustering result;

FIG. 9 is a schematic diagram of a training process of an ETA personalized model provided by an embodiment of the present application;

FIG. 10 is a flow chart of an ETA estimation method provided by an embodiment of the present application;

FIG. 11 is a block diagram of an ETA model training apparatus according to an embodiment of the present application;

FIG. 12 is a block diagram of an ETA model training apparatus according to another embodiment of the present application;

FIG. 13 is a block diagram of an ETA estimation device according to an embodiment of the present application;

FIG. 14 is a block diagram of a computer device provided in one embodiment of the present application;

fig. 15 is a block diagram of a terminal according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

In the technical scheme provided by the embodiment of the application, a concept of 'speed ratio' is provided. The speed ratio of the target user refers to a ratio of the moving speed of the target user to the road passing speed.

The speed ratio of the target user on a certain road refers to the ratio of the moving speed of the target user on the road to the road passing speed corresponding to the road. The road traffic speed is used for reflecting the moving speed of most users on the road. For example, the road traffic speed may be represented by an average or weighted average of the moving speeds of all users on the road. Therefore, the speed ratio of the target user on a certain road reflects the difference between the moving speed of the target user on the road and the moving speed of most users on the road, namely the personalized behavior habit of the target user on the road is quantitatively represented.

The speed ratio of the target user on the road with a certain attribute refers to a ratio of the moving speed of the target user on the road with the attribute to the road passing speed corresponding to the road with the attribute. Therefore, the speed ratio of the target user on the road with a certain attribute reflects the difference between the moving speed of the target user on the road with the attribute and the moving speed of most users on the road with the attribute, namely the personalized behavior habit of the target user on the road with the attribute is quantitatively represented. The attribute of the road is used to reflect the characteristics or state of the road. For example, the attributes of roads are divided in advance according to factors such as road grade, road congestion degree, road type, number of lanes, and road width.

According to the technical scheme provided by the embodiment of the application, the speed ratio characteristics of a target user are obtained by obtaining the speed ratio of the target user on various roads with different attributes, and then an ETA personalized model for pre-estimating the ETA of the target user is trained according to the speed ratio characteristics of the target user; the speed ratio characteristics can reflect the difference between the moving speed of the target user on the roads with different attributes and the moving speed of most users, namely the personalized behavior habits of the target user on the roads with different attributes are quantitatively represented, so that the ETA personalized model obtained by final training is more accurate in ETA estimation of the target user.

In addition, the technical scheme provided by the embodiment of the application is suitable for ETA estimation in any scenes such as driving, riding and walking. In the following method embodiments, the description is mainly given by taking a driving scene as an example, but the scope of protection of the present application is not limited.

In the following, the technical solution of the present application will be described in two aspects, namely, a training process and an application process of the ETA model.

In the embodiment of the training method related to the ETA model described below, the execution subject of each step may be a computer device. The Computer device refers to an electronic device with computing and processing capabilities, such as a PC (Personal Computer), a server, and the like.

In the following embodiments of the method for applying the ETA model (i.e., the ETA estimation method), the execution subject of each step may be a terminal device such as a mobile phone, a navigation device, and a vehicle-mounted terminal.

Referring to fig. 1, a flowchart of a method for training an ETA model according to an embodiment of the present application is shown. The method may comprise the steps of:

step 101, at least one piece of track data of a target user is obtained, wherein the at least one piece of track data corresponds to a road with at least one attribute.

Optionally, n pieces of trajectory data of the target user are obtained, where the n pieces of trajectory data correspond to roads of the a types of attributes, n is a positive integer, and a is a positive integer.

The target user may be any one of the users. The trajectory data of the target user is used to characterize the trajectory generated by the target user's movements. For example, a piece of trajectory data may include location information collected and recorded at predetermined intervals during movement from one geographic location to another.

In one possible implementation, the trajectory data of the target user is obtained by mining the navigation data of the target user. For example, 100 navigations are used by the target user, and 100 pieces of trajectory data of the target user can be obtained accordingly. In other possible implementation manners, the real-time geographical position of the target user may also be tracked and positioned in the moving process of the target user, so as to obtain the trajectory data of the target user.

In the embodiment of the present application, the attribute of the road is used to reflect the feature or state of the road. For example, the attributes of the roads are divided in advance according to the factors such as road grade, road congestion degree, road type, number of lanes, road width, etc., and the attributes are divided into 56 different attributes in total as shown in fig. 2.

One piece of trajectory data may include a road of the same attribute, or may include a plurality of roads of different attributes. For example, a certain piece of track data includes 2 roads with different attributes, one of which is a clear road and the other is a congested road. For another example, a certain piece of track data includes 3 roads with different attributes, where one is a clear and 3-lane road, another is a clear and 2-lane road, and yet another is a congested and 2-lane road. A "road" referred to herein may be referred to as a "link," which represents a segment of a road.

And integrating the n pieces of track data of the target user to obtain roads corresponding to the attributes of the a types. For example, including clear roads, slow roads, congested roads, elevated roads, tunneled roads, 1 lane roads, 2 lane roads, 3 lane roads, 4 lane roads, and so forth. The n pieces of trajectory data of the target user may or may not include one or more roads having a certain attribute. For example, the n pieces of trajectory data of the target user include 60 clear roads, 30 slow roads, 10 congested roads, and 0 extremely congested roads (i.e., no extremely congested roads).

And 102, acquiring the speed ratio of the target user on the road with the at least one attribute according to the at least one piece of track data.

Optionally, the speed ratio of the target user on the road with the a-type attributes is obtained according to the n pieces of trajectory data.

In the embodiment of the present application, the speed ratio of the target user on the road of the target attribute in the at least one attribute refers to a ratio of a moving speed of the target user on the road of the target attribute to a road passing speed corresponding to the road of the target attribute. For example, the speed ratio of the target user on the i-th attribute road in the a-type attributes is a ratio of the moving speed of the target user on the i-th attribute road to the road passing speed corresponding to the i-th attribute road, and i is a positive integer less than or equal to a.

The moving speed of the target user is different for different application scenes. For example, for a driving scene, the moving speed of the target user is the driving speed; for a riding scene, the moving speed of the target user is the riding speed; for a walking scene, the moving speed of the target user is the walking speed.

The road traffic speed is used for reflecting the moving speed of most users on the road. For example, the road traffic speed may be represented by an average or weighted average of the moving speeds of all users on the road.

The speed ratio of the target user on the road of the target attribute reflects the difference between the moving speed of the target user on the road of the target attribute and the moving speed of most users on the road of the target attribute. The larger the speed ratio of the target user on the road of the target attribute is, the faster the target user moves on the road of the target attribute than the average level is, and the user may be more accustomed to moving on the road of the attribute (such as driving); conversely, a smaller speed ratio of the target user on the road of the target attribute indicates that the target user moves slower on the road of the target attribute than on average, and the user may be less accustomed to moving (e.g., driving) on the road of the attribute.

Taking the target attribute as clear as an example, the speed ratio of the target user on the clear road refers to the ratio of the moving speed of the target user on the clear road to the road passing speed corresponding to the clear road. The speed ratio of the target user on the clear road reflects the difference between the moving speed of the target user on the clear road and the moving speed of most users on the clear road. The larger the speed ratio of the target user on the smooth road is, the higher the moving speed of the target user on the smooth road is compared with the average level; conversely, a smaller speed ratio of the target user on the clear road indicates that the target user moves slower on the clear road than on the average.

Alternatively, taking the example of calculating the speed ratio of the target user on the road of the target attribute, the calculation process includes the following sub-steps:

1. and selecting the track data corresponding to the target attribute from the at least one piece of track data for the target attribute in the at least one attribute.

Optionally, for the ith attribute of the a attributes, selecting the track data corresponding to the ith attribute from the n track data. Taking the ith attribute as clear as an example, the track data containing clear roads is selected from the n track data. For example, when n is 100, there are 80 pieces of trajectory data including a clear road.

2. And acquiring the speed ratio of the target user on at least one road with the target attribute according to the track data corresponding to the target attribute.

Optionally, the speed ratio of the target user on the m roads with the ith attribute is obtained according to the trajectory data corresponding to the ith attribute.

In the embodiment of the present application, the speed ratio of the target user on the target road refers to a ratio of a moving speed of the target user on the target road to a road passing speed corresponding to the target road. For example, the speed ratio of the target user on the k-th road of the m roads refers to a ratio of a moving speed of the target user on the k-th road to a road passing speed corresponding to the k-th road, where k is a positive integer and is less than or equal to m.

Still referring to the above example, assuming that a total of 200 clear road paths are included in 80 pieces of trajectory data including clear roads, the moving speed of the target user on each clear road and the road passing speed corresponding to the clear road path are respectively obtained, and the speed ratio of the target user on each clear road is calculated.

3. And obtaining the speed ratio of the target user on the road with the target attribute by weighted average according to the speed ratio of the target user on the road with the at least one target attribute and the length of the road with the at least one target attribute.

Optionally, the speed ratio of the target user on the road with the ith attribute is obtained by weighted average according to the speed ratio of the target user on the m roads and the lengths of the m roads.

For example, the speed ratio R of the target user on the road of the i-th attribute is calculated using the following formula_i：

Wherein R is_fIs the speed ratio, L, of the target user on the f-th road of the m roads_fIs the length of the f-th road of the m roads,

is the sum of the lengths of the m roads, and f is a positive integer less than or equal to m.

By the method, the speed ratio of the target user on the ith attribute road is calculated by adopting a weighted average algorithm according to the speed ratio of the target user on the m ith attribute roads and the lengths of the m roads, so that the calculation result of the speed ratio is more accurate.

Step 103, training an ETA personalized model for estimating the ETA of the target user according to the speed ratio characteristics of the target user.

In the embodiment of the present application, the speed profile of the target user includes the speed ratio of the target user on the road with at least one attribute acquired in the above step 102. Alternatively, it is assumed that the speed ratios of the target user on the roads with the a types of attributes are obtained in step 102, the speed ratio of the target user is characterized by the speed ratios of the target user on the roads with the total r types of attributes including the a type of attribute, and r is an integer greater than or equal to a.

If the n pieces of trajectory data of the target user obtained in step 101 include a road with a certain attribute, the speed ratio of the target user on the road with the attribute may be calculated by adopting the manner provided in step 102. For example, if the n pieces of trajectory data of the target user acquired in step 101 include a 3-lane road, the speed ratio of the target user on the 3-lane road may be calculated in the manner provided in step 102.

If there is no road with a certain attribute in the n pieces of trajectory data of the target user obtained in step 101, the speed ratio of the target user on the road with the attribute cannot be calculated in the manner provided in step 102. For example, if the n pieces of trajectory data of the target user acquired in step 101 do not include a 4-lane road, the speed ratio of the target user on the 4-lane road cannot be calculated in the manner provided in step 102. For this situation, a speed ratio filling mode may be adopted to obtain the speed ratio of the target user on the road of the attribute, which may be specifically described in the following embodiments.

In a possible case, a ═ r, that is, the road with all r attributes is included in the n pieces of trajectory data of the target user acquired in the above step 101, then for each attribute of the r attributes, the speed ratio of the target user on the road with the attribute can be calculated by adopting the manner provided in the above step 102. The r attribute is preset, and may be, for example, 56 attributes as shown in fig. 2.

In addition, the characteristics used for training the ETA personalized model can include route characteristics and user characteristics. Route characteristics may include characteristics such as physical attributes of the route (such as attributes of length, width, road grade, etc. of the route), historical mining speed of the route, real-time speed of the route, etc. Different from the related technology introduced in the background art, in the embodiment of the present application, the user characteristics include speed ratio characteristics of the target user, and the speed ratio characteristics can reflect the difference between the moving speed of the target user on the roads with different attributes and the moving speed of most users, that is, the personalized behavior habits of the target user on the roads with different attributes are quantitatively represented, so that the ETA model obtained through final training is more accurate when ETA estimation is performed on the target user.

It should be noted that, with the function of estimating the ETA of the target user provided by the ETA personalized model in the embodiment of the present application, the expected arrival time of the target user at the target location may be estimated, and the time duration consumed by the target user to reach the target location may also be estimated.

To sum up, in the technical scheme provided in the embodiment of the present application, speed ratio characteristics of a target user are obtained by obtaining speed ratios of the target user on various roads with different attributes, and then an ETA personalized model for estimating ETA of the target user is trained according to the speed ratio characteristics of the target user; on one hand, the speed bit characteristics can be obtained by automatically mining the track data of the user by means of data analysis and processing, so that the defects of long time consumption and difficulty in ensuring accuracy existing in manual mining are overcome; on the other hand, the speed ratio characteristics can reflect the difference between the moving speed of the target user on the roads with different attributes and the moving speed of most users, namely the personalized behavior habits of the target user on the roads with different attributes are quantitatively represented, so that the ETA personalized model obtained by final training is more accurate when ETA estimation is carried out on the target user.

As described above, if the speed ratio of the target user on the road with a certain attribute cannot be calculated according to the trajectory data of the target user, the speed ratio of the target user on the road with the attribute may be obtained by using a speed ratio filling method. In one example, assuming that the speed ratio of the target user on the road with the a types of attributes is calculated according to the n pieces of trajectory data of the target user by using the method provided in the embodiment of fig. 1, if a is smaller than r, the speed ratio characteristic of the target user further includes: the speed ratio of the target user on the road of at least one other attribute (in the embodiment of the present application, b is a positive integer) than the above-mentioned attribute a. The speed ratio on the road with the b other attributes can be obtained by adopting a speed ratio filling mode, and the process can comprise the following steps:

first, at least one similar user of the target user is obtained.

Optionally, s similar users of the target user are obtained, where s is a positive integer.

The similar users of the target user refer to users selected from the user set and similar to the personalized behavior habits of the target user. The number of similar users, that is, the value of s, may be preset, for example, s is 10, which is not limited in this embodiment of the application. In addition, the set of users includes the target user and at least one other user.

Optionally, this step comprises the following substeps:

1. and for the target attribute in the at least one attribute, determining the behavior distribution characteristics of the target user on the road of the target attribute according to the speed ratio of the target user on the road of the target attribute and the speed ratios of other users in the user set on the road of the target attribute.

In the embodiment of the present application, the behavior distribution feature of the target user on the road of the target attribute refers to a distribution of a speed ratio of the target user on the road of the target attribute in a speed ratio of each user in the user set on the road of the target attribute.

And the behavior distribution characteristic of the target user on the road of the target attribute is used for describing the behavior of the target user on the road of the target attribute, and the position of all users in the user set in the behavior distribution of the target attribute on the road. The behavior of the target user on the road of the target attribute may be characterized by a speed ratio of the target user on the road of the target attribute.

The determination process of the behavior distribution characteristics of the target users on the road with the target attributes is a discretization of the speed ratio of the target users on the road with the target attributes.

The behavior distribution of all users in the user set on the road of the target attribute refers to a probability distribution of speed ratios of all users in the user set on the road of the target attribute.

As shown in fig. 3, which is a schematic diagram illustrating a probability distribution of speed ratios of all users in a user set on a road of a certain attribute. Assuming that the target user has 3 pieces of trajectory data on the road of the attribute, and accordingly, 3 speed ratios of the target user on the road of the attribute are calculated, which are assumed to be 1.1, 1.5 and 0.8, the speed ratios of the target user on the road of the attribute, where all users in the user set are located in the speed ratio distribution on the road of the attribute, may be as shown in fig. 3.

Optionally, the behavior distribution characteristics of the target user on the road of the target attribute are determined by the following steps:

1.1, creating at least two sections according to the speed ratio of the target user on the road of the target attribute and the speed ratios of other users in the user set on the road of the target attribute.

The speed ratios of all users in the user set on the road of the target attribute range from [0, + ∞). If the speed ratio of a certain user on the road with the target attribute is 0, it indicates that the trajectory data of the user does not exist on the road with the target attribute, that is, the speed ratio of the user on the road with the target attribute cannot be calculated in the manner described in the embodiment of fig. 1 above.

Optionally, u intervals are created, u being an integer greater than 1. The value of u may be preset, for example, u is 10, that is, 10 intervals are created. In addition, each of the u sections corresponds to a range of values of one speed ratio.

In one possible embodiment, an average value of the speed ratios of all users in the user set on the road of the target attribute is calculated (assuming that the average value is expressed by mean), and at least one section is created on each of the left and right sides centering on the mean value mean. Where the section size may be a standard deviation of the speed ratios of all users in the user set on the road of the target attribute (assuming that the standard deviation is expressed by std).

In one example, as shown in fig. 4, 5 intervals are created on the left and right sides of the average mean, and the value ranges corresponding to the intervals are, in order from left to right: the value ranges corresponding to the 1 st interval are [0, mean-std 4 ], the value ranges corresponding to the 2 nd interval are [ mean-std 4, mean-std 3 ], the value ranges corresponding to the 3 rd interval are [ mean-std 3, mean-std 2 ], the value ranges corresponding to the 4 th interval are [ mean-std 2, mean-std 1 ], the value ranges corresponding to the 5 th interval are [ mean-std 1, mean ], the value ranges corresponding to the 6 th interval are [ mean, mean + std 1 ], the value ranges corresponding to the 7 th interval are [ mean + std 1, mean + std 2 ], the value ranges corresponding to the 8 th interval are [ mean + std 2, mean + std 3 ], the value ranges corresponding to the 8 th interval are [ mean + std 3, mean + std 4 th interval are [ mean + std 3 ], + ∞).

It should be noted that the number and size of the above-described intervals are exemplary and explanatory. In some other possible embodiments, the mean value mean may not be used as a center, such as a predetermined value; alternatively, the standard deviation std may not be used as the interval size, such as a preset value. In addition, the more the interval is divided, the more data support is provided based on the statistical characteristics, accordingly, the behavior distribution characteristics of the target user on the road with the ith attribute are more accurate, but the calculation amount is increased correspondingly, so that the factors of the precision and the calculation amount can be comprehensively balanced to determine the number of the intervals.

1.2, acquiring the section of the speed ratio of the target user on the road with at least one target attribute.

Optionally, the section to which the speed ratios of the target user on the m roads with the ith attribute belong respectively is obtained. Each interval has a corresponding value range, and after the speed ratios of the target user on the m roads with the ith attribute are obtained, the interval in which each speed ratio falls is determined respectively.

In one example, assume that for the ith attribute, a total of 4 intervals are created, respectively [0, 0.5), [0.5, 1), [1, 1.5), and [1.5, + ∞). Further, it is assumed that the 6 speed ratios of the target user on the road of the i-th attribute are obtained, which are 0.8, 0.9, 0.8, 1.1, 1.5, and 1.4, respectively. Then 3 of them fall within the interval [0.5, 1 ], the other 2 fall within the interval [1, 1.5 ] and the remaining 1 fall within the interval [1.5, + ∞ ] for the 1.5 speed ratios 1.1 and 1.4.

And 1.3, integrating the average values of the speed ratios of the target users in each of the at least two sections to obtain the behavior distribution characteristics of the target users on the road with the target attributes.

Optionally, the average value of the speed ratios of the target users respectively included in each of the u sections is integrated to obtain the behavior distribution characteristics of the target users on the road with the i-th attribute.

With reference to the above example, the average value of the speed ratios of the target users included in the section [0.5, 1) is (0.8+0.9+0.8)/3 ═ 0.83, the average value of the speed ratios of the target users included in the section [1, 1.5) is (1.1+1.4)/2 ═ 1.25, and the average value of the speed ratios of the target users included in the section [1.5, + ∞) is 1.5. In addition, since there is no speed ratio falling within the interval [0, 0.5) among the 6 speed ratios of the target user, the speed ratio of the target user included in the interval [0, 0.5) may be represented by a preset parameter indicating a speed ratio not falling within the interval.

For each attribute in the attributes a, the behavior distribution characteristics of the target user on the road with the attribute can be determined in the manner described in the above steps 1.1-1.3. In addition, the division manner and the number of the intervals corresponding to different attributes may be the same or different, and this is not limited in this application.

2. And determining the subject description characteristics of the target user according to the behavior distribution characteristics of the target user on the roads with the specified number of attributes.

Optionally, when the specified number is r, determining the topic description feature of the target user according to the behavior distribution feature of the target user on the roads with r types of attributes.

In the embodiment of the application, the topic description feature of the target user includes the probability that the target user belongs to each topic in t topics, and t is a positive integer. The number of themes may be predetermined, and the embodiment of the present application is not limited thereto.

In addition, for a attributes of the total r attributes, the behavior distribution characteristics of the target user on the road of each of the a attributes may be determined in the manner described above. When a is smaller than r, the behavior distribution characteristics of the target user on the roads with b other attributes besides the above-mentioned attribute a can be marked as null (i.e. indicating absence) or indicated in a preset manner. As shown in fig. 5, the behavior distribution characteristics of the target user on the r-attribute road are obtained, fig. 5 only shows the behavior distribution characteristics of the target user on the i-th attribute road, and the behavior distribution characteristics of the target user on the other attribute roads are similar to this.

In one possible implementation, the topic description feature of the target user can be obtained by training the topic model, and the process may include the following sub-steps:

2.1, training a theme model according to the behavior distribution characteristics of the target user on the road with the specified number of attributes and the behavior distribution characteristics of other users in the user set on the road with the specified number of attributes;

optionally, when the specified number is r, the topic model is trained according to the behavior distribution characteristics of the target user on the r-attribute roads and the behavior distribution characteristics of other users in the user set on the r-attribute roads.

The behavior distribution characteristics of the other users in the user set on the r-attribute roads can be obtained by referring to the above-described manner of determining the behavior distribution characteristics of the target user on the r-attribute roads.

A topic model (topic model) is a statistical model used for discovering abstract topics in a series of documents in the fields of machine learning, natural language processing and the like. In the embodiment of the application, abstract topics are found in a user set by means of a topic model, and topic description characteristics of each user in the user set are determined.

Optionally, the topic model adopts an LDA (Latent Dirichlet Allocation) model. In the embodiment of the application, behavior distribution characteristics of different users in the user set can be learned through the LDA model, so that the latent topic information in the user set can be identified. The LDA model is typically applied in the field of natural language processing, and is characterized by tf-idf (term frequency-inverse document frequency index) vectors of words in a dictionary or variants thereof. The LDA model is used for personalized user theme description of the ETA model in the traffic field, and the difficulty is in the design of ETA behavior distribution characteristics. The determination process of the behavior distribution characteristics provided by the application, no matter the meaning of the determination process or the application of the determination process in the LDA solving process, conforms to the essential idea of the LDA model.

Of course, in some other possible embodiments, the topic model may also adopt an LSA (Latent Semantic index) model, a pLSA (Latent Semantic index of probability) model, and the like, which is not limited in this application.

After behavior distribution characteristics of all users in the user set on the r-attribute roads are obtained, the number of the topics in the user set is specified, and a topic model is trained.

And 2.2, obtaining the theme description characteristics of the target user through the theme model.

Assuming that the number of topics is t, the topic description feature of each user t dimension in the user set (including the topic description feature of the target user t dimension) can be obtained through the trained topic model.

The t-dimension theme description features of the target user are normalized to obtain the probability that the target user belongs to each theme in the t themes, the value range of the probability belonging to any theme can be (0, 1), and the sum of the probabilities belonging to each theme in the t themes is 1.

As shown in fig. 6, the distribution diagram of the theme description features of a certain user is exemplarily shown, taking the number of themes as 20 as an example.

In addition, according to the trained topic model, the composition of the behavior distribution characteristics in each topic can be obtained through a visualization method. Each topic has a relatively independent behavior distribution characteristic, and the meaning of the topic is definite. Different topics describe different categories of users.

As shown in fig. 7, a schematic diagram of behavior distribution characteristics corresponding to 20 subjects is exemplarily shown. The behavior distribution characteristics in each theme are arranged in descending order according to the importance degree, and the importance degree is also a parameter output by the theme model and used for measuring the importance degree of the behavior distribution characteristics in the theme.

For example, in the topic 1, the moving speed of the user on the road having the attribute of the main and sub road connection road (formwaylroad 5) is greater than the road moving speed, the moving speed on the road having the attribute of the ramp (formwaylroad 11) is also greater than the road moving speed, the moving speed on the road having the attribute of the extreme congestion (jam) is significantly higher than the road moving speed, and the moving speed on the road having the attribute of the other road (road _ class8) is substantially identical to the road moving speed.

For another example, in topic 3, the moving speed of the user on a road having an attribute of 4 lanes (lane4) substantially coincides with the road moving speed, the moving speed on a road having an attribute of 150 breadth (width150) substantially coincides with the road moving speed, the moving speed on a road having an attribute of national road (road _ class2) is slightly higher than the road moving speed, and the moving speed on a road having an attribute of right turn (formwayRoad18) is slightly higher than the road moving speed.

For another example, in the topic 15, the moving speed of the user on the road having the attribute of the highway (road _ class0) is slightly slower than the road moving speed, the moving speed on the road having the attribute of JCT (a connecting road between a high speed and a high speed, formwaylroad 3) is substantially identical to the road moving speed, the moving speed on the road having the attribute of the fully enclosed road (formwaylroad 12) is substantially identical to the road moving speed, and the moving speed on the road having the attribute of the main road (formwaylroad 31) is slightly faster than the road moving speed.

3. And selecting at least one similar user of the target user from the user set according to the theme description characteristics of the target user and the theme description characteristics of other users in the user set.

Optionally, calculating similarity between the topic description features of the target user and other users, and selecting similar users of the target user based on the similarity, the process may include the following sub-steps:

and 3.1, calculating the similarity between the theme description characteristics of the target user and the theme description characteristics of other users in the user set.

In the embodiment of the application, the similarity between users is converted into the similarity between the theme description features of the users. The more similar the subject matter description characteristics of two users indicate that the two users are more similar. Optionally, the similarity between the user's topic description features is calculated by a distance function. The similarity can be expressed by any one or more combinations of weighted distance, Euclidean distance and cosine distance.

In one possible implementation, the similarity is represented in terms of a weighted distance. The weighted distance and the similarity are in a negative correlation relationship, and the smaller the weighted distance is, the higher the similarity is; conversely, the greater the weighted distance, the lower the similarity. Optionally, the weighted distance between the topic description features of the two users is calculated using the following formula:

weighted distance ═ euclidean distance · (1-cosine distance) + maximum distance;

the maximum distance is the distance of the dimension with the maximum Euclidean distance in the theme description characteristics of the t dimensions of the two users. The similarity is characterized by using the weighted distance, so that the method is more accurate.

Optionally, before calculating the similarity, clustering may be performed on the users in the user set to reduce the calculation amount of the similarity. Thus, this step may comprise several sub-steps:

a. and clustering all users in the user set.

Based on the topic description characteristics of each user in the user set, all users in the user set are clustered by adopting a related clustering algorithm to obtain a plurality of classes. Clustering algorithms include, but are not limited to, any of the following: a k-means algorithm, a DBSCAN (Density-based Clustering of Applications with Noise) algorithm, a Spectral Clustering (Spectral Clustering) algorithm.

b. And calculating the similarity between the subject description characteristics of the target user and the subject description characteristics of other users in the class to which the target user belongs.

As shown in fig. 8, all users in the user set are clustered, assuming 25 classes. Because the theme description feature is a probability feature, and the sum of the probabilities in three dimensions is 1, the clustering image is a plane with uneven density in three-dimensional coordinates. Assuming that the target user belongs to the 10 th class of the above 25 classes, the similarity between the topic description feature of the target user and the topic description feature of each other user in the 10 th class is calculated respectively. For example, if 100 users are included in the 10 th class, the similarity between the topic description feature of the target user and the topic description features of the remaining 99 other users is calculated respectively.

After clustering, only the similarity between the target user and other users in the class to which the target user belongs needs to be calculated, and the similarity between the target user and other users in the whole user set does not need to be calculated, so that the calculation amount is obviously reduced, and the efficiency of determining similar users is improved.

And 3.2, sorting according to the similarity from large to small, and selecting at least one other user which is sorted at the front in the sequence as at least one similar user of the target user.

Optionally, assuming that the number of similar users to be selected is s, s other users in the sequence are selected as s similar users of the target user.

Still referring to the above example, after the similarity between the target user and the 99 other users is obtained through calculation, the 99 other users are ranked in the order of decreasing similarity. Assuming s is 10, the first 10 other users in the sequence are selected as 10 similar users of the target user.

Secondly, determining the speed ratio of the target user on the road with at least one other attribute according to the speed ratio of at least one similar user on the road with at least one other attribute.

Optionally, the speed ratio of the target user on the road with the b other attributes is determined according to the speed ratios of the s similar users on the road with the b other attributes.

Optionally, for the j-th attribute of the b other attributes, the speed ratio of the target user on the road with the j-th attribute is obtained by weighted summation according to the speed ratio of the s similar users on the road with the j-th attribute and the similarity between the target user and the s similar users. The speed ratios of the s similar users on the road with the jth attribute respectively correspond to weights, the similarity between the s similar users and the target user is in positive correlation, and j is a positive integer less than or equal to b. The greater the similarity between the similar user and the target user is, the greater the weight corresponding to the speed ratio of the similar user on the road with the jth attribute is; conversely, the smaller the similarity between the similar user and the target user is, the smaller the weight corresponding to the speed ratio of the similar user on the road with the jth attribute is.

For the s similar users, the speed ratios of the similar users on the road with the jth attribute are sequentially searched according to the sequence of the similarity from large to small, and if the currently searched similar users have the speed ratios on the road with the jth attribute (that is, the similar users have the speed ratios calculated by adopting the method provided by the step 102 in the embodiment of fig. 1), the similar users and the corresponding speed ratios thereof are recorded until a preset number of speed ratios are obtained or all the s similar users are traversed, and then the operation is stopped. And then, according to the recorded speed ratio and the similarity corresponding to the similar users, carrying out weighted summation to obtain a calculation result, and taking the calculation result as the speed ratio of the target user on the road with the jth attribute, thereby realizing speed ratio filling.

In summary, in the technical scheme provided in the embodiment of the present application, the speed ratios of the target user on the roads with various attributes are completely obtained by obtaining the similar users of the target user and filling the vacant speed ratios of the target user according to the speed ratios of the similar users, so that the finally trained ETA personalized model of the target user can be suitable for ETA estimation of the target user on the road with any attribute, and the reliability of the ETA personalized model is ensured. Therefore, even on the road which the target user history does not reach, the higher accuracy of ETA estimation can be ensured.

In addition, all users in the user set are clustered firstly, and similar users of the target user are searched after clustering, so that the calculation amount of similarity calculation is reduced.

In the speed ratio filling process, the theme description characteristics of the target user can be obtained, and the theme description characteristics also reflect the personalized behavior habits of the target user on roads with different attributes. Therefore, when the ETA personalized model of the target user is trained, besides the speed ratio characteristic of the target user, the topic description characteristic of the target user can be fused.

In one example, an ETA personalized model for estimating ETA of a target user is trained according to personalized features of the target user; the personalized features of the target user comprise: the speed ratio characteristics of the target user, and the subject description characteristics of the target user.

Certainly, the user characteristics used for training the ETA personalized model may include, in addition to the above personalized characteristics, other user characteristics such as the user's own attributes (e.g., age, gender, etc.), the familiarity of the user with the route, the main activity city range of the user, etc., and model training is performed by combining the user characteristics and the route characteristics (e.g., the physical attributes of the route, the historical mining speed of the route, the real-time speed of the route, etc.), so as to finally obtain the ETA personalized model.

In the embodiment of the application, the probability that the target user belongs to each topic in all topics can be obtained through normalization according to the topic description features obtained through topic model training, personalized behavior habits of the target user on roads with different attributes are reflected, the topic description features and the speed ratio features are fused with the ETA personalized model for training, and the accuracy of the ETA personalized model obtained through final training in ETA estimation of the target user can be further improved.

Referring to fig. 9, a schematic diagram of a training process of an ETA personalization model is exemplarily shown. The training process may be as follows:

1. and obtaining the speed ratio characteristics of the target user according to the trajectory data of the target user.

Assuming that the target user has n pieces of trajectory data corresponding to the roads of the a types of attributes, the speed ratio of the target user on the road of each of the a types of attributes can be obtained according to the trajectory data.

2. And obtaining the behavior distribution characteristics of the target user according to the speed ratio characteristics of the target user.

3. And training the LDA model according to the behavior distribution characteristics of the target user and the behavior distribution characteristics of other users in the user set.

4. And obtaining the theme description characteristics of the target user through the trained LDA model.

Each user in the user set has a corresponding topic description feature, and the topic description feature is a probability feature.

5. And clustering the users in the user set according to the theme description characteristics of the users in the user set.

For example, each user in the set of users is clustered using a k-means clustering algorithm.

6. And according to the clustering result, finding out similar users of the target user, and filling the speed ratio characteristics of the target user according to the speed ratio characteristics of the similar users.

And carrying out similarity calculation on the target user and other users in the category to which the target user belongs pairwise to obtain a most similar user list of the target user. And filling the speed ratio of the target user on the road with the non-driving attribute according to the speed ratio of the similar users in the most similar user list.

7. And combining the speed bit of the target user and the theme description characteristics to obtain the personalized characteristics of the target user, and training the ETA personalized model according to the personalized characteristics of the target user.

Optionally, the personalized features of the target user and the route features are fused, and an ETA personalized model is trained by using a regression method.

Please refer to fig. 10, which illustrates a flowchart of an ETA estimation method according to an embodiment of the present application. The method can be applied to terminals such as mobile phones, navigation equipment, vehicle-mounted terminals and the like. The method may comprise the steps of:

step 1001, the trip information of the target user is acquired.

The trip information may include a start point position and an end point position.

In a possible application scenario, the target user opens the navigation software installed in the mobile phone, and inputs or selects the starting point position and the ending point position in the navigation software. Subsequently, the navigation software provides the planned route from the starting position to the end position for the target user, and the ETA of the planned route is predicted.

In another possible application scenario, the target user inputs or selects a start point position and an end point position in the navigation device or the vehicle-mounted terminal. Subsequently, the navigation device or the vehicle-mounted terminal provides a planned route from the starting position to the end position for the target user, and the ETA of the planned route is estimated.

At step 1002, a planned route from a start location to an end location is determined.

The terminal may determine a planned route from the starting position to the ending position according to a preset route planning rule. The number of the planned routes may be one or multiple, and the embodiment of the present application does not limit this. If the route is multiple planned routes, the ETA corresponding to each planned route can be estimated subsequently.

In addition, the planned route may include a plurality of roads, i.e., a plurality of links.

Step 1003, obtaining route characteristics corresponding to the planned route and user characteristics of the target user.

The route characteristics corresponding to the planned route may include: physical attributes of the planned route (such as the length, width, road grade and other attributes of each link included in the planned route), historical mining speed of the planned route (such as the historical mining speed of each link included in the planned route), real-time speed of the planned route (such as the real-time speed of each link included in the planned route), and other characteristics.

In this embodiment of the present application, the user characteristics of the target user include speed ratio characteristics of the target user, the speed ratio characteristics of the target user include speed ratios of the target user on the roads with the specified number of attributes, and the speed ratio of the target user on the road with the target attribute in the specified number of attributes refers to a ratio of a moving speed of the target user on the road with the target attribute to a road traveling speed corresponding to the road with the target attribute. For the description of the speed ratio characteristics, reference is made to the above embodiments, which are not repeated herein.

Optionally, the user characteristics of the target user further comprise the subject description characteristics of the target user introduced above.

And 1004, invoking an ETA personalized model of the target user, and obtaining an ETA estimation result corresponding to the planned route through the ETA personalized model according to the route characteristics corresponding to the planned route and the user characteristics of the target user.

And taking the route characteristics corresponding to the planned route and the user characteristics of the target user as input parameters of the ETA personalized model, carrying out fusion processing on the two characteristics by the ETA personalized model, and outputting an ETA estimation result corresponding to the planned route.

Because different users have their own corresponding ETA personalized models, even in the same route, the ETA estimated results obtained by using the ETA personalized models corresponding to the two users may be different, which also conforms to the actual situation.

In addition, in the above method embodiment, only the terminal calls the ETA personalized model to perform ETA prediction as an example, in other possible embodiments, after the terminal acquires the travel information of the target user, the travel information of the target user can also be sent to the background server, the background server performs route planning and calls the ETA personalized model to perform ETA prediction, and the planned route and the corresponding ETA prediction result are fed back to the terminal; or after the terminal acquires the travel information of the target user, the terminal can also locally determine a planned route, then sends the identification information of the target user and the planned route to the background server, the background server calls the ETA personalized model to perform ETA estimation, and the ETA estimation result is fed back to the terminal.

In summary, in the technical scheme provided by the embodiment of the application, the speed ratio feature is automatically mined from the trajectory data of the user, the ETA estimation is performed by fusing the route feature and the user feature (including the speed ratio feature) through the ETA personalized model, and compared with the ETA estimation performed according to the manually mined feature value, the speed ratio feature can objectively and quantitatively represent the personalized behavior habits of the user on roads with different attributes, so that the finally obtained ETA estimation result is more accurate.

The following are embodiments of the apparatus of the present application that may be used to perform embodiments of the method of the present application. For details which are not disclosed in the embodiments of the apparatus of the present application, reference is made to the embodiments of the method of the present application.

Referring to fig. 11, a block diagram of an ETA model training apparatus according to an embodiment of the present application is shown. The device has the functions of realizing the method examples, and the functions can be realized by hardware or by hardware executing corresponding software. The device can be a terminal or be arranged on the terminal. The apparatus 1100 may include: data acquisition module 1110, speed ratio acquisition module 1120, and model training module 1130.

The data acquiring module 1110 is configured to acquire at least one piece of trajectory data of a target user, where the at least one piece of trajectory data corresponds to a road with at least one attribute.

The speed ratio obtaining module 1120 is further configured to obtain, according to the at least one piece of trajectory data, a speed ratio of the target user on the road with the at least one attribute; wherein the speed ratio is a ratio of the moving speed of the target user to the road passing speed.

A model training module 1130, configured to train, according to the speed ratio characteristic of the target user, an ETA personalized model for estimating ETA of the target user; wherein the speed bit characteristic of the target user comprises a speed ratio of the target user on the road of the at least one attribute.

In an alternative embodiment provided based on the embodiment of fig. 11, the speed ratio obtaining module 1120 is configured to:

for a target attribute in the at least one attribute, selecting track data corresponding to the target attribute from the at least one piece of track data;

acquiring the speed ratio of the target user on at least one road with the target attribute according to the track data corresponding to the target attribute; the speed ratio of the target user on the target road refers to a ratio of the moving speed of the target user on the target road to the road passing speed corresponding to the target road;

and obtaining the speed ratio of the target user on the road with the target attribute by weighted average according to the speed ratio of the target user on the road with the at least one target attribute and the length of the road with the at least one target attribute.

In another optional embodiment provided based on the embodiment of fig. 11 or any one of the optional embodiments above, the speed bit characteristic of the target user comprises a speed ratio of the target user on a road of a specified number of attributes, the specified number of attributes comprising the at least one attribute and at least one other attribute;

as shown in fig. 12, the apparatus 1100 further includes: a similar user acquisition module 1140 and a speed ratio padding module 1150.

The similar user obtaining module 1140 is configured to obtain at least one similar user of the target user.

The speed ratio padding module 1150 is configured to determine the speed ratio of the target user on the road with the at least one other attribute according to the speed ratio of the at least one similar user on the road with the at least one other attribute.

Optionally, the similar user acquiring module 1140 includes: a distribution characteristic determination unit 1141, a description characteristic determination unit 1142, and a similar user selection unit 1143.

The distribution feature determining unit 1141 is configured to, for a target attribute of the at least one attribute, determine a behavior distribution feature of the target user on the road of the target attribute according to a speed ratio of the target user on the road of the target attribute and speed ratios of other users in the user set on the road of the target attribute; the user set comprises the target user and at least one other user, and the behavior distribution feature of the target user on the road of the target attribute refers to the distribution of the speed ratio of the target user on the road of the target attribute in the speed ratio of each user in the user set on the road of the target attribute.

The description feature determining unit 1142 is configured to determine a topic description feature of the target user according to the behavior distribution feature of the target user on the road with the specified number of attributes; wherein the topic description feature of the target user comprises a probability that the target user belongs to each topic of at least one topic.

The similar user selecting unit 1143 is configured to select at least one similar user of the target user from the user set according to the theme description feature of the target user and the theme description features of other users in the user set.

Optionally, the distribution characteristic determining unit 1141 is configured to:

creating at least two sections according to the speed ratio of the target user on the road of the target attribute and the speed ratios of other users in the user set on the road of the target attribute; wherein each of the at least two intervals corresponds to a range of values of a speed ratio;

acquiring the interval of the speed ratio of the target user on at least one target attribute road; the speed ratio of the target user on the target road refers to a ratio of the moving speed of the target user on the target road to the road passing speed corresponding to the target road;

and integrating the average value of the speed ratios of the target users contained in each of the at least two sections to obtain the behavior distribution characteristics of the target users on the road with the target attributes.

Optionally, the description feature determining unit 1142 is configured to:

training a theme model according to the behavior distribution characteristics of the target user on the road with the specified number of attributes and the behavior distribution characteristics of other users in the user set on the road with the specified number of attributes;

and obtaining the theme description characteristics of the target user through the theme model.

Optionally, the similar user selecting unit 1143 includes: a similarity degree subunit 1143a and a similar user selection subunit 1143 b.

The similarity operator unit 1143a is configured to calculate similarities between the topic description features of the target user and the topic description features of other users in the user set.

The similar user selecting subunit 1143b is configured to select, according to the similarity degree, at least one other user ranked earlier in the sequence as at least one similar user of the target user.

Optionally, the similarity operator unit 1143a is configured to:

clustering each user in the user set;

and calculating the similarity between the subject description characteristics of the target user and the subject description characteristics of other users in the category to which the target user belongs.

In another optional embodiment provided based on the embodiment of fig. 12 or any one of the optional embodiments above, the model training module 1130 is configured to:

training an ETA personalized model for estimating the ETA of the target user according to the personalized features of the target user;

wherein the personalized features of the target user include: a speed ratio characteristic of the target user, and a subject description characteristic of the target user.

In another optional embodiment provided based on the embodiment of fig. 12 or any of the optional embodiments above, the speed ratio padding module 1150 is configured to:

for target other attributes in the at least one other attribute, weighting and summing to obtain the speed ratio of the target user on the road of the target other attributes according to the speed ratio of the at least one similar user on the road of the target other attributes and the similarity between the target user and the at least one similar user;

and the speed ratios of the at least one similar user on the roads with other attributes of the target respectively correspond to weights, and the similarity between the at least one similar user and the target user is in positive correlation.

Please refer to fig. 13, which illustrates a block diagram of an ETA estimation apparatus according to an embodiment of the present application. The device has the functions of realizing the method examples, and the functions can be realized by hardware or by hardware executing corresponding software. The device can be a terminal or be arranged on the terminal. The apparatus 1300 may include: a trip acquisition module 1310, a route planning module 1320, a feature acquisition module 1330, and an ETA estimation module 1340.

A trip obtaining module 1310, configured to obtain trip information of a target user, where the trip information includes a start point position and an end point position.

A route planning module 1320 for determining a planned route from the start location to the end location.

A feature obtaining module 1330, configured to obtain a route feature corresponding to the planned route and a user feature of the target user, where the user feature of the target user includes a speed ratio feature of the target user, the speed ratio feature of the target user includes a speed ratio of the target user on a road with a specified number of attributes, and the speed ratio is a ratio of a moving speed of the target user to a road passing speed.

The ETA estimation module 1340 is configured to invoke the ETA personalized model of the target user, and obtain an ETA estimation result corresponding to the planned route according to the route characteristic corresponding to the planned route and the user characteristic of the target user through the ETA personalized model.

It should be noted that, when the apparatus provided in the foregoing embodiment implements the functions thereof, only the division of the functional modules is illustrated, and in practical applications, the functions may be distributed by different functional modules according to needs, that is, the internal structure of the apparatus may be divided into different functional modules to implement all or part of the functions described above. In addition, the apparatus and method embodiments provided by the above embodiments belong to the same concept, and specific implementation processes thereof are described in the method embodiments for details, which are not described herein again.

Referring to fig. 14, a schematic structural diagram of a computer device 1400 according to an embodiment of the present application is shown. The computer apparatus 1400 may be used to implement the training method of the ETA model provided in the above embodiments. Specifically, the method comprises the following steps:

the computer device 1400 includes a Central Processing Unit (CPU)1401, a system memory 1404 including a Random Access Memory (RAM)1402 and a Read Only Memory (ROM)1403, and a system bus 1405 connecting the system memory 1404 and the central processing unit 1401. The computer device 1400 also includes a basic input/output system (I/O system) 1406 that facilitates transfer of information between devices within the computer, and a mass storage device 1407 for storing an operating system 1413, application programs 1414, and other program modules 1415.

The basic input/output system 1406 includes a display 1408 for displaying information and an input device 1409, such as a mouse, keyboard, etc., for user input of information. Wherein the display 1408 and input device 1409 are both connected to the central processing unit 1401 via an input-output controller 1410 connected to the system bus 1405. The basic input/output system 1406 may also include an input/output controller 1410 for receiving and processing input from a number of other devices, such as a keyboard, mouse, or electronic stylus. Similarly, input-output controller 1410 also provides output to a display screen, a printer, or other type of output device.

The mass storage device 1407 is connected to the central processing unit 1401 through a mass storage controller (not shown) connected to the system bus 1405. The mass storage device 1407 and its associated computer-readable media provide non-volatile storage for the computer device 1400. That is, the mass storage device 1407 may include a computer readable medium (not shown) such as a hard disk or CD-ROM drive.

Without loss of generality, the computer-readable media may comprise computer storage media and communication media. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes RAM, ROM, EPROM, EEPROM, flash memory or other solid state memory technology, CD-ROM, DVD, or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices. Of course, those skilled in the art will appreciate that the computer storage media is not limited to the foregoing. The system memory 1404 and mass storage device 1407 described above may collectively be referred to as memory.

According to various embodiments of the present application, the computer device 1400 may also operate as a remote computer connected to a network via a network, such as the Internet. That is, the computer device 1400 may be connected to the network 1412 through the network interface unit 1411 connected to the system bus 1405, or may be connected to other types of networks or remote computer systems (not shown) using the network interface unit 1411.

The memory also includes one or more programs stored in the memory and configured to be executed by one or more processors. The one or more programs include instructions for implementing a training method for the ETA model described above.

Referring to fig. 15, a block diagram of a terminal 1500 according to an embodiment of the present application is shown. The terminal 1500 may be an electronic device such as a mobile phone, a tablet computer, a navigation device, a vehicle-mounted terminal, and the like.

In general, terminal 1500 includes: a processor 1501 and memory 1502.

Processor 1501 may include one or more processing cores, such as a 4-core processor, an 8-core processor, or the like. The processor 1501 may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field Programmable Gate Array), and a PLA (Programmable Logic Array). Processor 1501 may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also referred to as a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor 1501 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, processor 1501 may also include an AI (Artificial Intelligence) processor for processing computational operations related to machine learning.

The memory 1502 may include one or more computer-readable storage media, which may be non-transitory. The memory 1502 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 1502 is used to store at least one instruction for execution by processor 1501 to implement the ETA estimation method provided by the method embodiments of the present application.

In some embodiments, the terminal 1500 may further include: a peripheral interface 1503 and at least one peripheral. The processor 1501, memory 1502, and peripheral interface 1503 may be connected by buses or signal lines. Various peripheral devices may be connected to peripheral interface 1503 via buses, signal lines, or circuit boards. Specifically, the peripheral device may include: at least one of a radio frequency circuit 1504, a display 1505, an audio circuit 1506, a positioning assembly 1507, and a power supply 1508.

Those skilled in the art will appreciate that the configuration shown in fig. 15 does not constitute a limitation of terminal 1500, and may include more or fewer components than shown, or some components may be combined, or a different arrangement of components may be employed.

In an example embodiment, there is also provided a computer device comprising a processor and a memory having stored therein at least one instruction, at least one program, set of codes, or set of instructions. The at least one instruction, at least one program, set of code, or set of instructions is configured to be executed by one or more processors to implement the training method or the ETA prediction method of the ETA model described above. Optionally, the computer device is a server, a PC or a terminal device such as a mobile phone, a navigation device, a vehicle terminal as introduced above.

In an exemplary embodiment, a computer readable storage medium is also provided, in which at least one instruction, at least one program, a set of codes, or a set of instructions is stored, which when executed by a processor of a computer device implements the above-mentioned training method or ETA prediction method of an ETA model.

Alternatively, the computer-readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

In an exemplary embodiment, a computer program product is also provided, which, when executed, is adapted to implement the above-described training method of the ETA model or the ETA estimation method.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The above description is only exemplary of the present application and should not be taken as limiting the present application, and any modifications, equivalents, improvements and the like that are made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method of training an ETA model, the method comprising:

training an ETA personalized model for estimating the ETA of the target user according to the route characteristics and the speed ratio characteristics of the target user; wherein the speed bit characteristics of the target user comprise a speed ratio of the target user on the road of the at least one attribute, and the route characteristics comprise at least one of a physical attribute, a historical mining speed, and a real-time speed of the road.

2. The method according to claim 1, wherein the obtaining the speed ratio of the target user on the road of the at least one attribute according to the at least one piece of trajectory data comprises:

acquiring the speed ratio of the target user on at least one road with the target attribute according to the track data corresponding to the target attribute;

3. The method of claim 1, wherein the speed bit characteristic of the target user comprises a speed ratio of the target user on a road of a specified number of attributes, the specified number of attributes comprising the at least one attribute and at least one other attribute;

before training an ETA personalized model for estimating the ETA of the target user according to the route characteristics and the speed ratio characteristics of the target user, the method further comprises the following steps:

acquiring at least one similar user of the target user;

and determining the speed ratio of the target user on the road of the at least one other attribute according to the speed ratio of the at least one similar user on the road of the at least one other attribute.

4. The method of claim 3, wherein the obtaining at least one similar user of the target user comprises:

for a target attribute in the at least one attribute, determining a behavior distribution characteristic of the target user on the road of the target attribute according to the speed ratio of the target user on the road of the target attribute and the speed ratios of other users in the user set on the road of the target attribute; the user set comprises the target user and at least one other user, and the behavior distribution characteristic of the target user on the road of the target attribute refers to the distribution of the speed ratio of the target user on the road of the target attribute in the speed ratio of each user in the user set on the road of the target attribute;

determining the subject description characteristics of the target user according to the behavior distribution characteristics of the target user on the road with the specified number of attributes; wherein the topic description characteristics of the target user comprise a probability that the target user belongs to each topic of at least one topic;

and selecting at least one similar user of the target user from the user set according to the theme description characteristics of the target user and the theme description characteristics of other users in the user set.

5. The method according to claim 4, wherein the determining the behavior distribution characteristics of the target user on the road of the target attribute according to the speed ratio of the target user on the road of the target attribute and the speed ratios of other users in the user set on the road of the target attribute comprises:

acquiring the interval of the speed ratio of the target user on at least one target attribute road;

6. The method according to claim 4, wherein the determining the topic description feature of the target user according to the behavior distribution feature of the target user on the road with the specified number of attributes comprises:

7. The method according to claim 4, wherein said selecting at least one similar user of the target user from the user set according to the topic description characteristics of the target user and the topic description characteristics of other users in the user set comprises:

calculating similarity between the topic description characteristics of the target user and the topic description characteristics of other users in the user set;

and selecting at least one other user with the top ranking in the sequence as at least one similar user of the target user according to the ranking from big to small of the similarity.

8. The method of claim 7, wherein the calculating the similarity between the topic description features of the target user and the topic description features of other users in the set of users comprises:

clustering each user in the user set;

9. The method of any one of claims 4 to 8, wherein training an ETA personalization model for estimating the ETA of the target user according to the route characteristics and the speed ratio characteristics of the target user comprises:

training an ETA personalized model for predicting ETA of the target user according to the route characteristics and the personalized characteristics of the target user;

10. The method of claim 3, wherein the determining the speed ratio of the target user on the road of the at least one other attribute according to the speed ratio of the at least one similar user on the road of the at least one other attribute comprises:

11. An Estimated Time of Arrival (ETA) estimation method, comprising:

determining a planned route from the start location to the end location;

obtaining route characteristics corresponding to the planned route and user characteristics of the target user, wherein the user characteristics of the target user comprise speed ratio characteristics of the target user, the speed ratio characteristics of the target user comprise speed ratios of the target user on roads with specified quantity of attributes, the speed ratio refers to a ratio of moving speed of the target user to road running speed, and the route characteristics comprise at least one of physical attributes, historical mining speed and real-time speed of the planned route;

12. An apparatus for training an ETA model, the apparatus comprising:

the model training module is used for training an ETA personalized model for pre-estimating the ETA of the target user according to the route characteristics and the speed ratio characteristics of the target user; wherein the speed bit characteristics of the target user comprise a speed ratio of the target user on the road of the at least one attribute, and the route characteristics comprise at least one of a physical attribute, a historical mining speed, and a real-time speed of the road.

13. An Estimated Time of Arrival (ETA) estimation apparatus, comprising:

a characteristic obtaining module, configured to obtain a route characteristic corresponding to the planned route and a user characteristic of the target user, where the user characteristic of the target user includes a speed ratio characteristic of the target user, the speed ratio characteristic of the target user includes a speed ratio of the target user on a road with a specified number of attributes, the speed ratio is a ratio of a moving speed of the target user to a road traveling speed, and the route characteristic includes at least one of a physical attribute, a historical mining speed, and a real-time speed of the planned route;

14. A computer device comprising a processor and a memory, the memory having stored therein at least one instruction, at least one program, set of codes, or set of instructions, which is loaded and executed by the processor to implement the method of any one of claims 1 to 10 or to implement the method of claim 11.

15. A computer readable storage medium having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the method of any one of claims 1 to 10 or to implement the method of claim 11.