Information pushing method and device

By obtaining multi-dimensional user feature data and evaluation models, we determine the conversion rate optimization value of users in logistics express services, solving the problems of low efficiency, high cost and poor effect of information push, and achieving accurate recommendations and user experience improvement.

CN119089031APending Publication Date: 2024-12-06BEIJING JINGDONG QIANSHITECHNOLOGY CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202310658290.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-06-05
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In logistics express delivery services, users will only receive recommended information when placing an order, resulting in low efficiency, high cost and poor effect of information push.

Method used

By obtaining multi-dimensional user feature data, combining the evaluation model to determine the conversion rate optimization value of the user when the recommendation information is pushed or not, it can achieve accurate recommendation and improve the efficiency and effect of information push.

Benefits of technology

It realizes accurate recommendations for information push, reduces the cost of information push, and improves user experience and push effects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089031A_ABST
    Figure CN119089031A_ABST
Patent Text Reader

Abstract

The invention discloses an information pushing method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the following steps: acquiring user information of a plurality of users; wherein the user information comprises multi-dimensional user feature data; according to the user information of the plurality of users, determining a corresponding conversion rate optimization value of each user when the recommendation information is pushed or not based on an evaluation model; comparing the conversion rate threshold value with the conversion rate optimization value to determine a target user and a recommendation information type corresponding to the target user; and determining corresponding target recommendation information according to the recommendation information type, and pushing the target recommendation information to a user terminal corresponding to the target user. According to the embodiment, the information recommendation efficiency is improved, the information recommendation cost is reduced, and the information recommendation effect is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an information pushing method and device. Background Art

[0002] With the development of Internet technology, how to expand Internet business through information push is a problem faced by many companies at present. Different from other Internet businesses with high user participation, such as online shopping business, even if users do not place orders, they will generate a large number of browsing, following, adding to shopping cart and other operations. Such businesses can make intelligent recommendations based on the information generated during user participation. For logistics and express delivery services, users generally click when they place orders. Therefore, there is a lack of data guidance when pushing corresponding recommended information (i.e. business marketing information, including new business, promotional business, etc.). The relevant technology mainly adopts the method of manual random push, which leads to low efficiency of information push, high cost of information push, and poor effect of information push (i.e. low conversion rate of information push). Summary of the invention

[0003] In view of this, an embodiment of the present invention provides an information push method and device, which can combine multi-dimensional user feature data and an evaluation model to determine the user's conversion rate optimization value when recommending information push or not, thereby achieving accurate recommendation, improving information recommendation efficiency, reducing information recommendation costs, and improving information recommendation effects and user experience.

[0004] To achieve the above object, according to one aspect of an embodiment of the present invention, there is provided an information push method, comprising:

[0005] Obtaining user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data, and recommendation information push record dimension feature data;

[0006] According to the user information of multiple users, the conversion rate optimization value corresponding to each user when the recommended information is pushed or not is determined based on the evaluation model; the target user and the recommended information type corresponding to the target user are determined according to the conversion rate threshold and the conversion rate optimization value;

[0007] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the target user.

[0008] Further, according to the user information of the plurality of users, the step of determining the conversion rate optimization value corresponding to each user when the recommended information is pushed or not based on the evaluation model includes:

[0009] According to the user information of the plurality of users, determining the conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the first evaluation model;

[0010] Determine, based on the user information of the plurality of users, a conversion rate improvement value corresponding to each user when the recommended information is pushed or not based on the second evaluation model;

[0011] The conversion rate gain value and the conversion rate improvement value are weighted to determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not.

[0012] Furthermore, the first evaluation model is a gain evaluation model; the step of determining the conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the user information of multiple users and the first evaluation model includes:

[0013] Determine, based on the gain evaluation model, a first conversion rate corresponding to each user when the recommendation information is pushed and a second conversion rate corresponding to each user when the recommendation information is not pushed according to the user information of the plurality of users;

[0014] The difference between the first conversion rate and the second conversion rate is calculated to determine a conversion gain value.

[0015] Furthermore, the second evaluation model is a decision tree evaluation model; the step of determining the conversion rate improvement value corresponding to each user when the recommended information is pushed or not based on the second evaluation model according to the user information of the plurality of users includes:

[0016] Determine, based on the decision tree evaluation model, a first estimated conversion rate corresponding to each user when the recommendation information is pushed and a second estimated conversion rate corresponding to when the recommendation information is not pushed according to the user information of the plurality of users;

[0017] A conversion rate improvement value is determined based on the first estimated conversion rate and the second estimated conversion rate.

[0018] Furthermore, after the step of obtaining user information of multiple users, the method further includes:

[0019] Multiple users are classified according to user information to determine different types of user groups.

[0020] Furthermore, after the step of determining the target user and the type of recommended information corresponding to the target user according to the conversion rate threshold and the conversion rate optimization value, the method further includes:

[0021] Determine the target user group corresponding to the target user;

[0022] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to each target user in the target user group.

[0023] Furthermore, the step of determining corresponding target recommendation information according to the recommendation information type and pushing the target recommendation information to a user terminal corresponding to the target user further includes:

[0024] With information push budget and user conversion rate as constraints, the constraints are solved based on machine learning algorithms, and target users are updated based on the solution results;

[0025] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the updated target user.

[0026] Furthermore, the steps of solving the constraint conditions based on the machine learning algorithm include:

[0027] Initialize the parameters of information push budget and user conversion rate, randomly generate an initial solution, and calculate the initial function value corresponding to the initial solution;

[0028] When the iteration number threshold is not reached, the loop is executed: a random perturbation value is determined in the feasible solution, a new solution is generated according to the random perturbation value, the corresponding objective function value is calculated according to the new solution, the objective function value is compared with the previous function value, and the previous solution is updated according to the comparison result;

[0029] Determine the solution corresponding to the number of iterations reached as the solution result.

[0030] According to another aspect of an embodiment of the present invention, there is provided an information push device, comprising:

[0031] An acquisition module is used to acquire user information of multiple users; wherein the user information includes user and multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data, and recommendation information push record dimension feature data;

[0032] A processing module is used to determine, based on the user information of multiple users and the evaluation model, the conversion rate optimization value corresponding to each user when the recommended information is pushed or not; and to determine the target user and the type of recommended information corresponding to the target user according to the conversion rate threshold and the conversion rate optimization value;

[0033] The push module is used to determine the corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to the user terminal corresponding to the target user.

[0034] According to another aspect of an embodiment of the present invention, there is provided an electronic device for information push, comprising:

[0035] one or more processors;

[0036] a storage device for storing one or more programs,

[0037] When one or more programs are executed by one or more processors, the one or more processors implement any of the above-mentioned information pushing methods.

[0038] According to another aspect of the embodiments of the present invention, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, any of the above-mentioned information pushing methods is implemented.

[0039] An embodiment of the above invention has the following advantages or beneficial effects: because the user information of multiple users is obtained; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data; according to the user information of multiple users, the corresponding conversion rate optimization value of each user when the recommendation information is pushed or not is determined based on the evaluation model; the target user and the recommendation information type corresponding to the target user are determined according to the conversion rate threshold and the conversion rate optimization value; the corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the target user. The technical means overcomes the technical problems of low information push efficiency, high information push cost and poor information push effect due to the use of manual random push in the related technology, and thus achieves the ability to combine multi-dimensional user feature data and evaluation models to determine the user's conversion rate optimization value when the recommendation information is pushed or not, thereby achieving accurate recommendation, improving information recommendation efficiency, reducing information recommendation costs, and improving information recommendation effects and user experience.

[0040] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0042] Figure 1 is a schematic diagram of the main process of an information push method provided according to an embodiment of the present invention;

[0043] Figure 2a is a schematic diagram of the main process of an information push method provided according to another embodiment of the present invention;

[0044] Figure 2b Shows Figure 2a A schematic diagram of determining a user group in the method shown;

[0045] Figure 2c Shows Figure 2a A training flow chart of the gain evaluation model in the method shown;

[0046] Figure 2d Shows Figure 2a An evaluation flow chart of the gain evaluation model in the method shown;

[0047] Figure 2e Shows Figure 2a A flowchart of the training of the decision tree evaluation model in the method shown;

[0048] Figure 2f Shows Figure 2a An evaluation flow chart of the decision tree evaluation model in the method shown;

[0049] Figure 2g Shows Figure 2a A flowchart of determining the final corresponding target user and recommended information according to the predicted samples in the method shown;

[0050] Figure 2h Shows Figure 2a The main flow chart of the method according to the simulated annealing algorithm;

[0051] Figure 3 is a schematic diagram of main modules of an information push device provided according to an embodiment of the present invention;

[0052] Figure 4 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0053] Figure 5 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0055] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution of the present invention are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0056] In order to count the usage of our products / services, we will aggregate, analyze and use the technically processed user data and share the processed statistical information with third parties. We will use secure encryption technology and other methods to ensure that the recipient of the information cannot re-identify a specific individual.

[0057] Figure 1 is a schematic diagram of the main process of the information push method provided according to an embodiment of the present invention; Figure 1 As shown, the information push method provided by the embodiment of the present invention mainly includes:

[0058] Step S101, obtaining user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data.

[0059] Specifically, the feature data of the user portrait dimension include static features such as user age, region, gender, occupation, etc.; the feature data of the user behavior dimension include data representing behavioral characteristics such as the user's recent access data (such as within the past week or the past month), active login time period, etc.; the feature data of the order log dimension refers to feature data of order dimensions such as user order category, order time, and order amount; the feature data of the recommendation information push record dimension refers to data such as the recommended information records pushed to the user in the recent period (such as within the past week or the past month).

[0060] Based on the multi-dimensional user feature data obtained, it is helpful to combine the evaluation model in the future to determine the optimized conversion rate value of each user when faced with the push of recommended information and the push of recommended information, thereby providing data support for information push, improving the efficiency and accuracy of information push, and reducing the cost of information push.

[0061] Further, according to an embodiment of the present invention, after the step of obtaining user information of multiple users, the method further includes:

[0062] Multiple users are classified according to user information to determine different types of user groups.

[0063] Exemplarily, according to a specific implementation of an embodiment of the present invention, based on different dimensions such as natural conversion rate, user line sensitivity (for logistics and distribution business), portrait tags, and user embedding vectors (user embedding vectors, which represent a series of user behaviors (such as web browsing, clicking on ads, etc.) in vector form), multiple users can be clustered / classified according to user information to determine user groups representing different types, such as high conversion rate user groups, user groups in specific regions, and business-specified user groups. Through the above settings, different types of user groups are determined based on the classification of multiple users, which helps to quickly determine the user group corresponding to the target user after the target user is determined, reducing the workload of determining the object to be pushed in information push, and further improving the efficiency of information push.

[0064] Step S102, according to the user information of multiple users, based on the evaluation model, determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not; compare the size relationship between the conversion rate threshold and the conversion rate optimization value to determine the target user and the recommended information type corresponding to the target user.

[0065] Specifically, based on user information and evaluation models, the conversion rate optimization value corresponding to each user when faced with whether to push recommended information is determined. The conversion rate optimization value represents the optimization effect of the conversion rate brought about by pushing recommended information to the user, and then combined with the conversion rate threshold, the target user and the type of recommended information corresponding to the target user are determined. Through the above settings, the target user with a good information push conversion rate and the type of recommended information that the target user is interested in are determined, which is used as a data guide to significantly improve the push effect and push accuracy of subsequent information push.

[0066] Further, according to an embodiment of the present invention, the step of determining the conversion rate optimization value corresponding to each user when the recommended information is pushed or not based on the evaluation model according to the user information of multiple users includes:

[0067] According to the user information of the plurality of users, determining the conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the first evaluation model;

[0068] Determine, based on the user information of the plurality of users, a conversion rate improvement value corresponding to each user when the recommended information is pushed or not based on the second evaluation model;

[0069] The conversion rate gain value and the conversion rate improvement value are weighted to determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not.

[0070] Exemplarily, according to an embodiment of the present invention, the first evaluation model is a gain evaluation model; the step of determining the conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the user information of the plurality of users and based on the first evaluation model comprises:

[0071] Determine, based on the gain evaluation model, a first conversion rate corresponding to each user when the recommendation information is pushed and a second conversion rate corresponding to each user when the recommendation information is not pushed according to the user information of the plurality of users;

[0072] The difference between the first conversion rate and the second conversion rate is calculated to determine a conversion gain value.

[0073] Among them, the gain evaluation model adopts an uplift model (gain model) based on T-learner (i.e., the Two Model dual model, which is to model the control group and the experimental group separately, and then the difference in the corresponding conversion rate between the experimental group model and the control group model is the uplift value (gain value)). The gain model can accurately determine the "incremental improvement" brought by whether or not the recommended information is pushed, thereby maximizing the conversion rate of the recommended information push and reducing the cost of information push.

[0074] Optionally, according to an embodiment of the present invention, the second evaluation model is a decision tree evaluation model; the step of determining, based on the user information of the plurality of users and based on the second evaluation model, the corresponding conversion rate improvement value for each user when the recommended information is pushed or not, comprises:

[0075] Determine, based on the decision tree evaluation model, a first estimated conversion rate corresponding to each user when the recommendation information is pushed and a second estimated conversion rate corresponding to when the recommendation information is not pushed according to the user information of the plurality of users;

[0076] A conversion rate improvement value is determined based on the first estimated conversion rate and the second estimated conversion rate.

[0077] The decision tree evaluation model uses the LightGBM model (an implementation of a gradient boosting decision tree, whose essential principle is to use the base classifier (decision tree) training integration to obtain the optimal model). The decision tree evaluation model is used to evaluate the estimated conversion rates of users when recommended information is pushed and when recommended information is not pushed, and then the conversion rate improvement value is determined by the difference between the two estimated conversion rates. Since the decision tree evaluation model scans all data samples for each input feature to evaluate the information gain of all possible branch points, it can take multi-dimensional user feature data into consideration to obtain a more accurate evaluation.

[0078] Through the above settings, the conversion rate gain value and conversion rate improvement value corresponding to whether the user pushes recommended information or not are determined by two modeling methods, namely the gain evaluation model and the decision tree evaluation model, and then the final conversion rate optimization value is determined by weighted fusion, which is used as the basis for determining the target users in turn, thereby improving the accuracy of the determined target users and improving the effect of subsequent information push.

[0079] Step S103: determining corresponding target recommendation information according to the recommendation information type, and pushing the target recommendation information to a user terminal corresponding to the target user.

[0080] Further, according to an embodiment of the present invention, after the step of determining the target user and the type of recommended information corresponding to the target user according to the conversion rate threshold and the conversion rate optimization value, the method further includes:

[0081] Determine the target user group corresponding to the target user;

[0082] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to each target user in the target user group.

[0083] Since different types of user groups have been determined through classification as described above, after determining the target users, in order to improve the efficiency of information push, the target user groups corresponding to the target users can be determined first, and then information push can be performed for batch users, so as to further improve the efficiency of information push while ensuring the push effect.

[0084] Preferably, according to an embodiment of the present invention, the step of determining corresponding target recommendation information according to the recommendation information type and pushing the target recommendation information to a user terminal corresponding to the target user further includes:

[0085] With information push budget and user conversion rate as constraints, the constraints are solved based on machine learning algorithms, and target users are updated based on the solution results;

[0086] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the updated target user.

[0087] The above provides another aspect of the information push method, which takes the information push budget into consideration, that is, constructs a method with information push budget and user conversion rate as constraints, screens target users according to machine learning algorithms (such as simulated annealing algorithm, and other existing algorithms can also be used to find local optimal solutions), and finally pushes information. Through the above settings, the cost of information push is significantly reduced, accurate information push is achieved, resource waste in the information push process is reduced, and user experience is improved.

[0088] According to the technical solution of the embodiment of the present invention, the user information of multiple users is obtained; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data; according to the user information of multiple users, the conversion rate optimization value corresponding to each user when the recommendation information is pushed or not is determined based on the evaluation model; the target user and the recommendation information type corresponding to the target user are determined according to the conversion rate threshold and the conversion rate optimization value; the corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the target user. Therefore, the technical problem of low information push efficiency, high information push cost and poor information push effect caused by the use of artificial random push in the related technology is overcome, and then the technical effect of being able to combine multi-dimensional user feature data and evaluation model to determine the conversion rate optimization value of the user when the recommendation information is pushed or not is achieved, thereby realizing accurate recommendation, improving information recommendation efficiency, reducing information recommendation cost, and improving information recommendation effect and user experience is achieved.

[0089] Figure 2a is a schematic diagram of the main process of the information push method provided according to another embodiment of the present invention; Figure 2a As shown, the information push method provided by the embodiment of the present invention mainly includes:

[0090] Step S201, obtaining user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data, and recommendation information push record dimension feature data.

[0091] Based on the multi-dimensional user feature data obtained, it is helpful to combine the evaluation model in the future to determine the optimized conversion rate value of each user when faced with the push of recommended information and the push of recommended information, thereby providing data support for information push, improving the efficiency and accuracy of information push, and reducing the cost of information push.

[0092] Step S202: classify multiple users according to the user information to determine different types of user groups.

[0093] Through the above settings, different types of user groups are determined based on the classification processing of multiple users, which helps to quickly determine the user group corresponding to the target user after the target user is determined, reduces the workload of determining the objects to be pushed for information push, and further improves the efficiency of information push.

[0094] Figure 2b A schematic diagram of determining multiple user groups is shown, such as Figure 2bAs shown, multi-dimensional user feature data of multiple users (user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data) are received. Based on different dimensions such as natural conversion rate, user line sensitivity (for logistics distribution business), portrait label, user embedding vector (user embedding vector, a series of user behaviors (such as web browsing, clicking on ads) are represented in vector form) clustering, multiple users can be clustered / classified according to user information to determine user groups representing different types, such as high conversion rate user groups, specific area user groups, business designated user groups, etc.

[0095] Step S203, according to the user information of multiple users, based on the gain evaluation model, determine the first conversion rate corresponding to each user when the recommendation information is pushed and the second conversion rate corresponding to the recommendation information is not pushed; calculate the difference between the first conversion rate and the second conversion rate to determine the conversion rate gain value.

[0096] Among them, the gain evaluation model adopts an uplift model (gain model) based on T-learner (i.e., the Two Model dual model, which is to model the control group and the experimental group separately, and then the difference in the corresponding conversion rate between the experimental group model and the control group model is the uplift value (gain value)). The gain model can accurately determine the "incremental improvement" brought by whether or not the recommended information is pushed, thereby maximizing the conversion rate of the recommended information push and reducing the cost of information push.

[0097] Specifically, the calculation formula for determining the conversion rate optimization value corresponding to each user when the recommended information is pushed or not based on the gain evaluation model is as follows:

[0098] τ(X i )=p(Y i |X i ,T i =1)-p(Y i |X i ,T i =0)

[0099] Among them, T i =1 means push recommended information, T i =0 means no recommended information is pushed (i.e., recommended information is not pushed); p(Y i |X i ,T i =1) represents the first conversion rate when recommending information to user i, p(Y i |X i ,T i =0) The second conversion rate when no recommended information is pushed to user i, τ(X i) represents the conversion rate gain value corresponding to user i when the recommended information is pushed compared with when the recommended information is not pushed.

[0100] According to an embodiment of the present invention, the above-mentioned information push method also includes a training and evaluation process of a gain evaluation model, such as Figure 2c and Figure 2d As shown, it mainly includes that, for the training process, the user samples for pushing recommended information and the user samples for not pushing recommended information in the training group data are separately modeled, wherein the first model G T Used to learn the first conversion rate of users with push recommendation information, and the second model G C It is used to learn the second conversion rate of users without push recommendation information. The difference between the two is the Uplift value (i.e., the conversion rate gain value, i.e. Figure 2c U=G T -G C ). For the evaluation process, the model G is used in the prediction group data. T and Model G C To predict the user conversion rate of users under whether to push recommended information, the difference between the two model prediction results (i.e., the first conversion rate and the second conversion rate) is calculated to obtain the final conversion rate gain value τ(X) = P(T = 1|X i )-P(T=0|X i ), where T=1 indicates that recommended information is pushed, and T=0 indicates that recommended information is not pushed (i.e., recommended information is not pushed).

[0101] Step S204, based on the user information of multiple users, determine the first estimated conversion rate corresponding to each user when the recommendation information is pushed and the second estimated conversion rate corresponding to the recommendation information is not pushed based on the decision tree evaluation model; based on the first estimated conversion rate and the second estimated conversion rate, determine the conversion rate improvement value.

[0102] The decision tree evaluation model uses the LightGBM model (an implementation of a gradient boosting decision tree, whose essential principle is to use the base classifier (decision tree) training integration to obtain the optimal model). The decision tree evaluation model is used to evaluate the estimated conversion rates of users when recommended information is pushed and when recommended information is not pushed, and then the conversion rate improvement value is determined by the difference between the two estimated conversion rates. Since the decision tree evaluation model scans all data samples for each input feature to evaluate the information gain of all possible branch points, it can take multi-dimensional user feature data into consideration to obtain a more accurate evaluation.

[0103] According to an embodiment of the present invention, the above-mentioned information push method also includes a training and evaluation process of a decision tree evaluation model, such as Figure 2e and Figure 2fAs shown, during the training process of the decision tree evaluation model, for the training group data, the samples with pushed recommendation information and the samples without pushed recommendation information are separated, and the LightGBM estimation model for pushing recommendation information and the LightGBM estimation model for not pushing recommendation information are trained respectively. During the evaluation process of the decision tree evaluation model, for all samples in the prediction group data, samples with pushed recommendation information features and samples without pushed recommendation information features are constructed respectively. For these two types of samples, the LightGBM estimation model for pushing recommendation information and the LightGBM estimation model for not pushing recommendation information are used for evaluation and prediction respectively. Finally, the prediction results of the two (i.e., the first estimated conversion rate and the second estimated conversion rate) are subtracted to obtain the final conversion rate improvement value of the pushed recommendation information.

[0104] Step S205: weighting the conversion rate gain value and the conversion rate improvement value to determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not.

[0105] Through the above settings, the conversion rate gain value and conversion rate improvement value corresponding to whether the user pushes recommended information or not are determined by two modeling methods, namely the gain evaluation model and the decision tree evaluation model, and then the final conversion rate optimization value is determined by weighted fusion, which is used as the basis for determining the target users in turn, thereby improving the accuracy of the determined target users and improving the effect of subsequent information push.

[0106] Step S206: determine the target user group corresponding to the target user.

[0107] Since different types of user groups have been determined through classification as described above, after determining the target users, in order to improve the efficiency of information push, the target user groups corresponding to the target users can be determined first, and then information push can be performed for batch users, so as to further improve the efficiency of information push while ensuring the push effect.

[0108] Step S207, using the information push budget and user conversion rate as constraints, based on the simulated annealing algorithm, update the users included in the target user group; determine the corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to the user terminals corresponding to each user in the updated target user group.

[0109] The above provides another aspect of the information push method, which takes the information push budget into consideration, that is, constructs a constraint condition based on the information push budget and the user conversion rate, combines the simulated annealing algorithm to screen the target users, and finally pushes the information. Through the above setting, the cost of information push is significantly reduced, accurate information push is achieved, resource waste in the information push process is reduced, and user experience is improved.

[0110] like Figure 2gThe flowchart of determining the final corresponding target user and recommended information according to the predicted sample is shown; specifically, according to the embodiment of the present invention, the sample data is firstly input into the gain evaluation model and the decision tree evaluation model respectively to calculate the conversion rate gain value and the conversion rate improvement value corresponding to the user; then the conversion rate gain value and the conversion rate improvement value are weightedly fused to obtain the conversion rate optimization value corresponding to the user. Then the conversion rate optimization value and the conversion rate threshold are compared to determine the target user and the target recommended information.

[0111] According to an embodiment of the present invention, the optimization process for the push object (target user) based on optimization conditions such as budget is as follows:

[0112] The optimization objective can be:

[0113]

[0114]

[0115]

[0116]

[0117]

[0118] Among them, τ ik is the conversion rate gain value of user i under the effect of recommended information type k; p ik is the order rate of user i under the influence of recommendation information type k; A ik is the order amount of user i under the influence of recommendation information type k; W ik represents the cost of recommending information type k; X ik Indicates whether user i is pushed recommended information type k; Budget is the budget for information recommendation; ROI refers to the return-on-investment ratio. It should be noted that the above optimization objectives (constraints) are not intended to limit this application and can be adjusted according to actual conditions.

[0119] Specifically, after the constraints are determined, the constraints can be solved by a simulated annealing algorithm (it should be noted that the solution of the simulated annealing algorithm is not a limitation of the present application, and other machine learning algorithms in the existing methods can also be used to solve the constraints) to update the users included in the target user group. After the update, the corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to each user in the updated target user group, so as to further reduce the waste of resources in the information push process while ensuring the push efficiency and push effect, and improve the user experience.

[0120] Exemplarily, the above step of solving the constraint conditions based on the machine learning algorithm includes:

[0121] Initialize the parameters of information push budget and user conversion rate, randomly generate an initial solution, and calculate the initial function value corresponding to the initial solution; when the iteration number threshold is not reached, execute the loop: determine a random perturbation value in the feasible solution, generate a new solution according to the random perturbation value, calculate the corresponding objective function value according to the new solution, compare the objective function value with the previous function value, and update the previous solution according to the comparison result; determine the corresponding solution when the iteration number is reached as the solution result.

[0122] like Figure 2g As shown, a specific implementation of an embodiment of the present invention provides a step of solving the constraint condition using a simulated annealing algorithm. Based on the above, the parameters of the constraint condition indicating the information push budget and the user conversion rate are initialized, an initial solution x0 is randomly generated, and the initial function value f(x0) corresponding to the initial solution x0 is calculated. Then the solution process is iterated: a random disturbance value Δx is determined in the feasible solution, and x is set to ′ =x0+Δx, where x ′ That is the new solution, based on the new solution x ′ Calculate the objective function value f(x ′ ). Then, by comparing the objective function value f(x ′ ) and the previous function value f(x0) to determine whether the new solution is due to the previous solution. If so, the new solution is used as the current solution; if not, the Metropolis criterion is used to determine whether to accept the new solution. When the iteration number threshold is reached, the final solution is output.

[0123] According to the technical solution of the embodiment of the present invention, the user information of multiple users is obtained; wherein the user information includes user and multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data; according to the user information of multiple users, the conversion rate optimization value corresponding to each user when the recommendation information is pushed or not is determined based on the evaluation model; the target user and the recommendation information type corresponding to the target user are determined according to the conversion rate threshold and the conversion rate optimization value; the corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the target user. Therefore, the technical problem of low information push efficiency, high information push cost and poor information push effect caused by the use of artificial random push in the related technology is overcome, and then the technical effect of being able to combine multi-dimensional user feature data and evaluation model to determine the conversion rate optimization value of the user when the recommendation information is pushed or not is achieved, thereby realizing accurate recommendation, improving information recommendation efficiency, reducing information recommendation cost, and improving information recommendation effect and user experience is achieved.

[0124] Figure 3 is a schematic diagram of the main modules of the information push device 300 provided according to an embodiment of the present invention; Figure 3 As shown, the information push device 300 provided by the embodiment of the present invention mainly includes:

[0125] The acquisition module 301 is used to obtain user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data.

[0126] Specifically, the feature data of the user portrait dimension include static features such as user age, region, gender, occupation, etc.; the feature data of the user behavior dimension include data representing behavioral characteristics such as the user's recent access data (such as within the past week or the past month), active login time period, etc.; the feature data of the order log dimension refers to feature data of order dimensions such as user order category, order time, and order amount; the feature data of the recommendation information push record dimension refers to data such as the recommended information records pushed to the user in the recent period (such as within the past week or the past month).

[0127] Based on the multi-dimensional user feature data obtained, it is helpful to combine the evaluation model in the future to determine the optimized conversion rate value of each user when faced with the push of recommended information and the push of recommended information, thereby providing data support for information push, improving the efficiency and accuracy of information push, and reducing the cost of information push.

[0128] Further, according to an embodiment of the present invention, the information push device 300 further includes a classification module. After the step of obtaining user information of multiple users, the classification module is used to:

[0129] Multiple users are classified according to user information to determine different types of user groups.

[0130] Exemplarily, according to a specific implementation of an embodiment of the present invention, based on different dimensions such as natural conversion rate, user line sensitivity (for logistics and distribution business), portrait tags, and user embedding vectors (user embedding vectors, which represent a series of user behaviors (such as web browsing, clicking on ads, etc.) in vector form), multiple users can be clustered / classified according to user information to determine user groups representing different types, such as high conversion rate user groups, user groups in specific regions, and business-specified user groups. Through the above settings, different types of user groups are determined based on the classification of multiple users, which helps to quickly determine the user group corresponding to the target user after the target user is determined, reducing the workload of determining the object to be pushed in information push, and further improving the efficiency of information push.

[0131] Processing module 302 is used to determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not based on the user information of multiple users and the evaluation model; determine the target user and the recommended information type corresponding to the target user based on the conversion rate threshold and the conversion rate optimization value.

[0132] Specifically, based on the multi-dimensional user feature data and evaluation model in the user information, the conversion rate optimization value corresponding to each user when faced with whether to push recommended information is determined. The conversion rate optimization value represents the optimization effect of the conversion rate brought about by pushing recommended information to the user, and then combined with the conversion rate threshold, the target user and the type of recommended information corresponding to the target user are determined. Through the above settings, the target user with a good information push conversion rate and the type of recommended information that the target user is interested in are determined, which is used as a data guide to significantly improve the push effect and push accuracy of subsequent information push.

[0133] Further, according to an embodiment of the present invention, the processing module 302 is further configured to:

[0134] According to the user information of the plurality of users, determining the conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the first evaluation model;

[0135] Determine, based on the user information of the plurality of users, a conversion rate improvement value corresponding to each user when the recommended information is pushed or not based on the second evaluation model;

[0136] The conversion rate gain value and the conversion rate improvement value are weighted to determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not.

[0137] Exemplarily, according to an embodiment of the present invention, the first evaluation model is a gain evaluation model; the processing module 302 is further used for:

[0138] Determine, based on the gain evaluation model, a first conversion rate corresponding to each user when the recommendation information is pushed and a second conversion rate corresponding to each user when the recommendation information is not pushed according to the user information of the plurality of users;

[0139] The difference between the first conversion rate and the second conversion rate is calculated to determine a conversion gain value.

[0140] Among them, the gain evaluation model adopts an uplift model (gain model) based on T-learner (i.e., the Two Model dual model, which is to model the control group and the experimental group separately, and then the difference in the corresponding conversion rate between the experimental group model and the control group model is the uplift value (gain value)). The gain model can accurately determine the "incremental improvement" brought by whether or not the recommended information is pushed, thereby maximizing the conversion rate of the recommended information push and reducing the cost of information push.

[0141] Optionally, according to an embodiment of the present invention, the processing module 302 is further configured to:

[0142] Determine, based on the decision tree evaluation model, a first estimated conversion rate corresponding to each user when the recommendation information is pushed and a second estimated conversion rate corresponding to when the recommendation information is not pushed according to the user information of the plurality of users;

[0143] A conversion rate improvement value is determined based on the first estimated conversion rate and the second estimated conversion rate.

[0144] The decision tree evaluation model uses the LightGBM model (an implementation of a gradient boosting decision tree, whose essential principle is to use the base classifier (decision tree) training integration to obtain the optimal model). The decision tree evaluation model is used to evaluate the estimated conversion rates of users when recommended information is pushed and when recommended information is not pushed, and then the conversion rate improvement value is determined by the difference between the two estimated conversion rates. Since the decision tree evaluation model scans all data samples for each input feature to evaluate the information gain of all possible branch points, it can take multi-dimensional user feature data into consideration to obtain a more accurate evaluation.

[0145] Through the above settings, the conversion rate gain value and conversion rate improvement value corresponding to whether the user pushes recommended information or not are determined by two modeling methods, namely the gain evaluation model and the decision tree evaluation model, and then the final conversion rate optimization value is determined by weighted fusion, which is used as the basis for determining the target users in turn, thereby improving the accuracy of the determined target users and improving the effect of subsequent information push.

[0146] The push module 303 is used to determine corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to the user terminal corresponding to the target user.

[0147] Further, according to an embodiment of the present invention, after the step of determining the target user and the type of recommended information corresponding to the target user according to the conversion rate threshold and the conversion rate optimization value, the push module 303 is further used to:

[0148] Determine the target user group corresponding to the target user;

[0149] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to each target user in the target user group.

[0150] Since different types of user groups have been determined through classification as described above, after determining the target users, in order to improve the efficiency of information push, the target user groups corresponding to the target users can be determined first, and then information push can be performed for batch users, so as to further improve the efficiency of information push while ensuring the push effect.

[0151] Preferably, according to an embodiment of the present invention, the push module 303 is further used for:

[0152] With information push budget and user conversion rate as constraints, target users are updated based on simulated annealing algorithm;

[0153] The corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the updated target user.

[0154] The above provides another aspect of the information push device, which takes the information push budget into consideration, that is, constructs a constraint condition based on the information push budget and user conversion rate, screens the target users according to the machine learning algorithm (such as simulated annealing algorithm), and finally pushes the information. Through the above setting, the cost of information push is significantly reduced, accurate information push is achieved, resource waste in the information push process is reduced, and user experience is improved.

[0155] According to the technical solution of the embodiment of the present invention, the user information of multiple users is obtained; wherein the user information includes user and multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data; according to the user information of multiple users, the conversion rate optimization value corresponding to each user when the recommendation information is pushed or not is determined based on the evaluation model; the target user and the recommendation information type corresponding to the target user are determined according to the conversion rate threshold and the conversion rate optimization value; the corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the target user. Therefore, the technical problem of low information push efficiency, high information push cost and poor information push effect caused by the use of artificial random push in the related technology is overcome, and then the technical effect of being able to combine multi-dimensional user feature data and evaluation model to determine the conversion rate optimization value of the user when the recommendation information is pushed or not is achieved, thereby realizing accurate recommendation, improving information recommendation efficiency, reducing information recommendation cost, and improving information recommendation effect and user experience is achieved.

[0156] Figure 4 An exemplary system architecture 400 is shown to which the information push method or information push device according to the embodiment of the present invention can be applied.

[0157] like Figure 4As shown, system architecture 400 may include terminal devices 401, 402, 403, network 404 and server 405. Network 404 is used to provide a medium for communication links between terminal devices 401, 402, 403 and server 405. Network 404 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0158] Users can use terminal devices 401, 402, 403 to interact with server 405 through network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, 403, such as shopping applications, web browser applications, search applications, instant messaging tools, logistics and distribution clients / applets, social platform software, etc. (only for example).

[0159] The terminal devices 401 , 402 , and 403 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0160] The server 405 may be a server that provides various services, such as a server that (push information / process data) to the user using the terminal devices 401, 402, and 403 (only as an example). The server may analyze and process the received user information and other data, and feed back the processing results (such as target recommendation information, target user - only as an example) to the terminal device.

[0161] It should be noted that the information push method provided in the embodiment of the present invention is generally executed by the server 405 , and accordingly, the information push device is generally disposed in the server 405 .

[0162] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0163] Reference below Figure 5 , which shows a schematic diagram of the structure of a computer system 500 of a terminal device or a server suitable for implementing an embodiment of the present invention. Figure 5 The terminal device or server shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0164] like Figure 5As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage part 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the system 500 are also stored. The CPU 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0165] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, etc.; an output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed, so that a computer program read therefrom is installed into the storage section 508 as needed.

[0166] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or installed from the removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-mentioned functions defined in the system of the present invention are executed.

[0167] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0168] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0169] The modules involved in the embodiments of the present invention may be implemented in software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor includes an acquisition module, a processing module, and a push module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the acquisition module may also be described as a "module for acquiring user information of multiple users."

[0170] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: obtaining user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data, and recommendation information push record dimension feature data; according to the user information of multiple users, determine the corresponding conversion rate optimization value of each user when the recommendation information is pushed or not based on the evaluation model; determine the target user and the recommendation information type corresponding to the target user according to the conversion rate threshold and the conversion rate optimization value; determine the corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to the user terminal corresponding to the target user.

[0171] According to the technical solution of the embodiment of the present invention, the user information of multiple users is obtained; wherein the user information includes user and multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data and recommendation information push record dimension feature data; according to the user information of multiple users, the conversion rate optimization value corresponding to each user when the recommendation information is pushed or not is determined based on the evaluation model; the target user and the recommendation information type corresponding to the target user are determined according to the conversion rate threshold and the conversion rate optimization value; the corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to the user terminal corresponding to the target user. Therefore, the technical problem of low information push efficiency, high information push cost and poor information push effect caused by the use of artificial random push in the related technology is overcome, and then the technical effect of being able to combine multi-dimensional user feature data and evaluation model to determine the conversion rate optimization value of the user when the recommendation information is pushed or not is achieved, thereby realizing accurate recommendation, improving information recommendation efficiency, reducing information recommendation cost, and improving information recommendation effect and user experience is achieved.

[0172] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An information push method, characterized in that: include: Acquire user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data, and recommendation information push record dimension feature data; According to the user information of the multiple users, determining the conversion rate optimization value corresponding to each user when the recommended information is pushed or not based on the evaluation model; comparing the size relationship between the conversion rate threshold and the conversion rate optimization value to determine the target user and the recommended information type corresponding to the target user; Determine corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to a user terminal corresponding to the target user.

2. The information push method according to claim 1, characterized in that: The step of determining the conversion rate optimization value corresponding to each user when the recommended information is pushed or not based on the evaluation model according to the user information of the multiple users comprises: Determine, based on the user information of the plurality of users, a conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the first evaluation model; Determine, based on the user information of the plurality of users, a conversion rate improvement value corresponding to each user when the recommended information is pushed or not based on a second evaluation model; The conversion rate gain value and the conversion rate improvement value are weighted to determine the conversion rate optimization value corresponding to each user when the recommended information is pushed or not.

3. The information push method according to claim 2, characterized in that: The first evaluation model is a gain evaluation model; the step of determining the conversion rate gain value corresponding to each user when the recommended information is pushed or not based on the user information of the multiple users and the first evaluation model comprises: Determine, based on the gain evaluation model, a first conversion rate corresponding to each user when the recommendation information is pushed and a second conversion rate corresponding to each user when the recommendation information is not pushed according to the user information of the multiple users; The difference between the first conversion rate and the second conversion rate is calculated to determine the conversion rate gain value.

4. The information push method according to claim 2, characterized in that: The second evaluation model is a decision tree evaluation model; the step of determining, based on the user information of the multiple users and based on the second evaluation model, the conversion rate improvement value corresponding to each user when the recommended information is pushed or not, comprises: Determining, based on the decision tree evaluation model, a first estimated conversion rate corresponding to each user when the recommendation information is pushed and a second estimated conversion rate corresponding to when the recommendation information is not pushed according to the user information of the multiple users; The conversion rate improvement value is determined based on the first estimated conversion rate and the second estimated conversion rate.

5. The information push method according to claim 1, characterized in that: After the step of obtaining user information of a plurality of users, the method further includes: The multiple users are classified according to the user information to determine different types of user groups.

6. The information push method according to claim 5, characterized in that: After the step of determining the target user and the type of recommended information corresponding to the target user according to the conversion rate threshold and the conversion rate optimization value, the method further includes: Determine a target user group corresponding to the target user; Corresponding target recommendation information is determined according to the recommendation information type, and the target recommendation information is pushed to a user terminal corresponding to each target user in the target user group.

7. The information push method according to claim 1, characterized in that: The step of determining corresponding target recommendation information according to the recommendation information type and pushing the target recommendation information to a user terminal corresponding to the target user further includes: Taking the information push budget and the user conversion rate as constraints, solving the constraints based on the machine learning algorithm, and updating the target users according to the solution results; Determine corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to a user terminal corresponding to the updated target user.

8. The information push method according to claim 7, characterized in that: The step of solving the constraint conditions based on the machine learning algorithm comprises: Initializing the parameters of the information push budget and the user conversion rate, randomly generating an initial solution, and calculating an initial function value corresponding to the initial solution; When the iteration number threshold is not reached, the following steps are executed in a loop: determining a random perturbation value in a feasible solution, generating a new solution according to the random perturbation value, calculating a corresponding objective function value according to the new solution, comparing the objective function value with a previous function value, and updating the previous solution according to the comparison result; The solution corresponding to when the number of iterations is determined to be the solution result.

9. An information push device, characterized in that: include: An acquisition module, used to acquire user information of multiple users; wherein the user information includes multi-dimensional user feature data, and the multi-dimensional user feature data includes user portrait dimension feature data, user behavior dimension feature data, order log dimension feature data, and recommendation information push record dimension feature data; A processing module, configured to determine, based on the user information of the plurality of users and based on the evaluation model, a conversion rate optimization value corresponding to each user when the recommended information is pushed or not; and compare the size relationship between the conversion rate threshold and the conversion rate optimization value to determine a target user and a recommended information type corresponding to the target user; The push module is used to determine corresponding target recommendation information according to the recommendation information type, and push the target recommendation information to a user terminal corresponding to the target user.

10. An electronic device for information push, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

11. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Citation Information

Patent Citations

  • Method of estimating advertisement conversion rate and servers

    CN108230010A

  • Information recommendation method and device

    CN113761344A

  • Information recommendation method and device and storage medium

    CN115222433A

  • Industrial digital twin systems and methods with echelons of executive, advisory and operations messaging and visualization

    US20220108262A1

  • Information recommendation method and apparatus, and electronic device and storage medium

    WO2020232615A1