Resource recommendation method, device, computer equipment and storage medium

By acquiring and processing interactive and display information of resources, combining historical data, calculating target smoothing interaction factors and matching user attributes, the problem of low effectiveness in traditional resource recommendation is solved, achieving more accurate resource recommendation and reducing resource waste.

CN114254181BActive Publication Date: 2024-11-08TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202010996171.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-09-21
Publication Date
2024-11-08
Estimated Expiration
2040-09-21

AI Technical Summary

Technical Problem

Traditional resource recommendation methods have low effectiveness, leading users to repeatedly search or refresh the interface, wasting computer resources.

Method used

By acquiring the target interaction information and target display information of the resources to be recommended, and combining them with historical interaction information and display information, sample smoothing is performed to calculate the target smoothing interaction factor. Based on the attribute information of the target users, information matching is performed to recommend resources.

Benefits of technology

It improves the effectiveness of resource recommendations, reduces repeated user searches and interface refreshes, and saves terminal or server resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114254181B_ABST
    Figure CN114254181B_ABST
Patent Text Reader

Abstract

The present application relates to a resource recommendation method, device, computer equipment and storage medium. The method comprises: obtaining target interaction information and target display information of a resource to be recommended; obtaining historical interaction information and historical display information of a historically recommended resource; determining a target interaction factor and a target display factor based on the target interaction information, target display information, historical interaction information and historical display information; performing sample smoothing processing according to the target interaction factor and the target display factor to obtain a target smoothed interaction factor corresponding to the resource to be recommended; obtaining attribute information of a target user, performing information matching based on the attribute information and the target smoothed interaction factor corresponding to the resource to be recommended to obtain a matching degree between the resource to be recommended and the target user; and recommending the resource to be recommended according to the matching degree. The use of this method can improve the effectiveness of resource recommendation and reduce the waste of computer equipment resources.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a resource recommendation method, device, computer equipment and storage medium. Background Art

[0002] With the development of computer technology, more and more people communicate and obtain various resources through network applications, for example, browsing news through social applications, browsing fund information through financial applications, etc. In order to facilitate users' viewing, network applications can actively recommend resources to users.

[0003] In traditional technologies, resources are usually recommended to users randomly, resulting in the recommended resources often not being of interest to users, and the effectiveness of resource recommendations is low. Summary of the invention

[0004] Based on this, it is necessary to provide a resource recommendation method, apparatus, computer device and storage medium that can improve the effectiveness of resource recommendation in response to the above technical problems.

[0005] A resource recommendation method, the method comprising:

[0006] Obtain target interaction information and target display information of the resource to be recommended;

[0007] Obtain historical interaction information and historical display information of historical recommended resources;

[0008] Determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information, and the historical display information;

[0009] Perform sample smoothing according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended;

[0010] Acquire the attribute information of the target user, perform information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtain the matching degree between the resource to be recommended and the target user;

[0011] Recommend resources based on matching degree.

[0012] A resource recommendation device, comprising:

[0013] An information acquisition module is used to acquire target interaction information and target display information of the resource to be recommended; and to acquire historical interaction information and historical display information of the historically recommended resource;

[0014] A factor determination module, used to determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information;

[0015] An information smoothing module, used to perform sample smoothing processing according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended;

[0016] A matching degree determination module is used to obtain attribute information of a target user, and to perform information matching based on the attribute information and a target smooth interaction factor corresponding to the resource to be recommended, so as to obtain a matching degree between the resource to be recommended and the target user;

[0017] A resource recommendation module is used to recommend the resource to be recommended according to the matching degree.

[0018] In one embodiment, the information acquisition module is further used to acquire resource categories of resources to be recommended; acquire reference interaction information and reference display information corresponding to the resource categories; use the reference interaction information as target interaction information, and use the reference display information as target display information.

[0019] In one embodiment, the information acquisition module is also used to obtain resource attribute information of the resources to be recommended; input the resource attribute information into a trained interaction factor prediction model to obtain a reference interaction factor; the interaction factor prediction model is trained based on the resource attribute information of the training recommended resources and the training interaction factor; obtain the resource category of the resources to be recommended, obtain the reference display information corresponding to the resource category, and use the reference display information as the target display information; obtain the target interaction information based on the reference interaction factor and the reference display information.

[0020] In one embodiment, the factor determination module is also used to obtain interactive fusion information based on the target interactive information and the historical interactive information; obtain display fusion information based on the target display information and the historical display information; use the ratio of the interactive fusion information to the display fusion information as the target interactive factor, and use the display fusion information as the target display factor.

[0021] In one embodiment, the information smoothing module is also used to obtain a first preset smoothing factor; scale the target display factor according to the first preset smoothing factor to obtain a target scaled display factor; and fuse the target interaction factor according to the target scaled display factor and the target display factor to obtain a target smoothed interaction factor.

[0022] In one embodiment, the target interaction factor includes a first interaction factor and a second interaction factor, the target display factor includes a first display factor and a second display factor, and the factor determination module is further used to use the ratio of the target interaction information to the target display information as the first interaction factor, and the target display information as the first display factor; use the ratio of the historical interaction information to the historical display information as the second interaction factor, and use the historical display information as the second display factor.

[0023] In one embodiment, the information smoothing module is also used to obtain a second preset smoothing factor; scale the first display factor according to the second preset smoothing factor to obtain a first scaled display factor; fuse the first interaction factor according to the first scaled display factor and the first display factor to obtain a first smoothed interaction factor; scale the second display factor according to the second preset smoothing factor to obtain a second scaled display factor; fuse the second interaction factor according to the second scaled display factor and the second display factor to obtain a second smoothed interaction factor; obtain a target smoothed interaction factor according to the first smoothed interaction factor and the second smoothed interaction factor.

[0024] In one embodiment, the information smoothing module is also used to obtain a first proportional adjustment factor corresponding to the first smooth interaction factor, and to obtain a second proportional adjustment factor corresponding to the second smooth interaction factor; the first smooth interaction factor and the second smooth interaction factor are adjusted according to the first proportional adjustment factor and the second proportional adjustment factor to obtain a target smooth interaction factor.

[0025] In one embodiment, the information smoothing module is further used to obtain the confidence interval and expectation corresponding to the first smooth interaction factor; obtain the first proportional adjustment factor according to the distance between the confidence interval and the target expectation; and determine the second proportional adjustment factor based on the first proportional adjustment factor.

[0026] In one embodiment, the information smoothing module is further used to obtain a confidence interval corresponding to the first smooth interaction factor; calculate the width of the confidence interval to obtain a first proportional adjustment factor; and determine a second proportional adjustment factor based on the first proportional adjustment factor.

[0027] In one embodiment, the information smoothing module is further used to obtain an interaction influence factor corresponding to the first interaction factor; adjust the first proportional adjustment factor according to the interaction influence factor; and determine the second proportional adjustment factor based on the adjusted first proportional adjustment factor.

[0028] In one embodiment, the matching degree determination module is further used to obtain a trained information matching model; input the attribute information and the target smooth interaction factor into the trained information matching model to obtain the matching degree between the resource to be recommended and the target user.

[0029] In one embodiment, the information acquisition module is also used to obtain training samples of the information matching model to be trained; the training samples include the training smooth interaction factor corresponding to the training recommendation resource, the attribute information of the training user, and the matching labels of the training recommendation resource and the training user; the training smooth interaction factor corresponding to the training recommendation resource and the attribute information of the training user are used as inputs of the information matching model to obtain the predicted matching results of the training recommendation resource and the training user; the model parameters of the information matching model are adjusted based on the predicted matching results and the matching labels until the convergence conditions are met to obtain the trained information matching model.

[0030] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0031] Obtain target interaction information and target display information of the resource to be recommended;

[0032] Obtain historical interaction information and historical display information of historical recommended resources;

[0033] Determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information, and the historical display information;

[0034] Perform sample smoothing according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended;

[0035] Acquire the attribute information of the target user, perform information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtain the matching degree between the resource to be recommended and the target user;

[0036] Recommend resources based on matching degree.

[0037] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0038] Obtain target interaction information and target display information of the resource to be recommended;

[0039] Obtain historical interaction information and historical display information of historical recommended resources;

[0040] Determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information, and the historical display information;

[0041] Perform sample smoothing according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended;

[0042] Acquire the attribute information of the target user, perform information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtain the matching degree between the resource to be recommended and the target user;

[0043] Recommend resources based on matching degree.

[0044] The resource recommendation method, device, computer equipment and storage medium described above obtain target interaction information and target display information of the resource to be recommended; obtain historical interaction information and historical display information of the historically recommended resource; determine the target interaction factor and target display factor based on the target interaction information, target display information, historical interaction information and historical display information; perform sample smoothing processing according to the target interaction factor and target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended; obtain the attribute information of the target user, perform information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtain the matching degree between the resource to be recommended and the target user; and recommend the resource to be recommended according to the matching degree. In this way, the calculation of the target smooth interaction factor of the resource to be recommended comprehensively considers its own interaction information and display information as well as the interaction information and display information of the historically recommended resource, so that the target smooth interaction factor can better reflect the characteristics of the resource to be recommended. The resource to be recommended is recommended to the target user according to the matching degree, and the calculation of the matching degree comprehensively considers the attribute information of the target user and the target smooth interaction factor of the resource to be recommended, so that the recommendation of the resource is more in line with the characteristics of the user, the recommendation is more accurate, and the effectiveness of the resource recommendation is improved. Effective resource recommendations can avoid users from repeatedly searching or refreshing the interface due to invalid resource recommendations. Repeated searches or repeated refreshes of the interface will take up a lot of computer device resources. Therefore, on the basis of improving the effectiveness of resource recommendations, it can also reduce the waste of terminal or server resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 An application environment diagram of a resource recommendation method in an embodiment;

[0046] Figure 2 A schematic diagram of a process flow of a resource recommendation method in one embodiment;

[0047] Figure 3 A schematic diagram of a flow chart of a resource recommendation method in another embodiment;

[0048] Figure 4 A schematic diagram of a flow chart of a resource recommendation method in another embodiment;

[0049] Figure 5 A schematic diagram of an upper bound, a lower bound, and an expected change curve of a confidence interval corresponding to a smooth interaction factor in an embodiment;

[0050] Figure 6 is a schematic diagram of a change curve corresponding to a weighting factor in an embodiment;

[0051] Figure 7 A schematic diagram showing the comparison of samples before and after smoothing in one embodiment;

[0052] Figure 8This is a schematic diagram of a fund recommendation interface in an embodiment;

[0053] Fig. 9 This is a schematic diagram of a news recommendation interface in one embodiment;

[0054] Fig.10 is a structural block diagram of a resource recommendation device in an embodiment;

[0055] Fig.11 is an internal structure diagram of a computer device in one embodiment;

[0056] Fig.12 FIG. 4 is a diagram showing the internal structure of a computer device in another embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] The resource recommendation method provided in this application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the server 104 through a network. The terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices, and the server 104 can be implemented as an independent server or a server cluster consisting of multiple servers.

[0059] The terminal 102 and the server 104 can be used alone to execute the resource recommendation method provided in the embodiment of the present application. The terminal 102 and the server 104 can also be used together to execute the resource recommendation method provided in the embodiment of the present application.

[0060] In one embodiment, Figure 2 As shown, a resource recommendation method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate, and the computer device can be the above Figure 1 The terminal 102 or the server 104 in the resource recommendation method includes the following steps:

[0061] Step S202: obtaining target interaction information and target display information of the resource to be recommended.

[0062] Step S204: Obtain historical interaction information and historical display information of historically recommended resources.

[0063] Among them, the resource to be recommended refers to the resource used to recommend to the target user. Resource refers to information that can be obtained through the network application, including at least one of text information, picture information, video information and audio information. For example, if the network application is an e-commerce application, the resource can be various merchants and various commodities. If the network application is a video application, the resource can be various videos. If the network application is a live broadcast application, the resource can be various live broadcast rooms. If the network application is a financial application, the resource can be various funds. If the network application is a social application, the resource can be various news information and friend dynamics. If the network application is a music application, the resource can be various songs. The resource to be recommended can be a resource that has been released for a period of time or a newly released resource. Historically recommended resources refer to resources that have been released for a period of time, which can be resources that have been recommended to the target user or resources that have not been recommended to the target user. The resource to be recommended and the historically recommended resources can be resources of the same resource theme, for example, the resource to be recommended and the historically recommended resources are both funds, and the resource to be recommended and the historically recommended resources are both videos. Furthermore, the resources to be recommended and the historically recommended resources can be resources of the same resource category. For example, the resources to be recommended and the historically recommended resources are both technology funds, and the resources to be recommended and the historically recommended resources are both character pictures. It can be understood that resource category is a more detailed word than resource theme. For example, if the resource theme is fund, the resource category can be technology fund, life fund, etc. If the resource theme is music, the resource category can be classical music, rock music, etc. If the resource theme is news, the resource category can be entertainment news, technology news, etc.

[0064] Interactive information refers to information generated through user interaction with resources. For example, interactive information may be the number of clicks, purchases, comments, and forwardings of resources. Target interactive information refers to interactive information of resources to be recommended, and historical interactive information refers to interactive information of historically recommended resources. Display information refers to information generated through resource display. For example, display information may be the amount of exposure of resources. Target display information refers to display information of resources to be recommended, and historical display information refers to display information of historically recommended resources. Historically recommended resources include at least one resource. When historically recommended resources include multiple resources, historical interactive information of historically recommended resources is the statistical value of interactive information of multiple resources, and historical display information of historical resources is the statistical value of display information of multiple resources. The statistical value of interactive information may be the average value, median value, etc. of interactive information. The statistical value of display information may be the average value, median value, etc. of display information.

[0065] Specifically, the computer device may count the interactive related information and display related information of the historical recommended resources within a preset time period, and obtain the historical interactive information and historical display information of the historical recommended resources. The preset time period may be a historical time period pre-set as needed, and the historical time period refers to the time period before the current moment. The unit of the historical time period may be a day, a month, a year, etc. For example, the number of clicks and exposures of the historical recommended fund in April this year are counted. When the resource to be recommended is a resource that has been released for a period of time, the computer device may count the interactive related information and display related information of the resource to be recommended within the same preset time period, and obtain the target interactive information and target display information of the resource to be recommended. For example, the number of clicks and exposures of the fund to be recommended and the historical recommended fund in April this year are counted. When the resource to be recommended is a newly released resource, the target interactive information and target display information of the resource to be recommended may be zero. When the resource to be recommended is a newly released resource, the computer device may also obtain the statistical values ​​of the historical interactive information and historical display information of multiple historical recommended resources belonging to the same category as the resource to be recommended as the target interactive information and target display information. For example, if the resource to be recommended is a newly released beauty product, the average number of clicks and average exposures of multiple released beauty products within a preset time period can be obtained as the number of clicks and exposures of the resource to be recommended. If the resource to be recommended is a newly released technology fund, the average number of clicks and average exposures of multiple released technology funds within a preset time period can be obtained as the number of clicks and exposures of the resource to be recommended. When the resource to be recommended is a newly released resource, the computer device can also obtain the statistical value of the historical display information of multiple historical recommended resources of the same category as the resource to be recommended as the target display information, and predict the target interaction information of the fund to be recommended through the model and the target display information.

[0066] Step S206, determining a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information.

[0067] The target interaction factor is the interaction ratio calculated based on the interaction information and the display information, and the target display factor is the display statistics calculated based on the display information. The display statistics can be the sum, weighted sum, etc. of the display information.

[0068] Specifically, the computer device can calculate the ratio of the target interactive information to the target display information to obtain a first interactive factor, calculate the ratio of the historical interactive information to the historical display information to obtain a second interactive factor, use the first interactive factor and the second interactive factor as the target interactive factor, and use the target display information and the historical display information as the target display factor. For example, if the resource to be recommended is the published fund A, and the historical recommended resources are other published funds, the click-through rate of fund A can be calculated based on the click volume and exposure volume of fund A in a preset time period, and the click-through rate of fund A can be used as the target interactive factor, and the exposure volume of fund A in the preset time period can be used as the target display factor. The average click-through rate is calculated based on the average click volume and average exposure volume of other published funds, and the average click-through rate is used as the target interactive factor, and the average exposure volume is used as the target display factor. The resource to be recommended is the published news B, and the historical recommended resources are other published news of the same category as news B. The comment rate of news B can be calculated based on the number of comments and exposure of news B in a preset time period, and the comment rate of news B is used as the target interaction factor, and the exposure of news B in the preset time period is used as the target display factor. The average comment rate can be calculated based on the average number of comments and average exposure of other published news of the same category as news B, and the average comment rate is used as the target interaction factor, and the average exposure is used as the target display factor.

[0069] The computer device may also calculate the target interaction information, target display information, and the comprehensive statistical value of the historical interaction information and historical display information of the resource to be recommended as the target interaction factor, and calculate the comprehensive value of the target display information of the resource to be recommended and the historical display information of the historically recommended resource as the target display factor. For example, if the resource to be recommended is the released fund A, and the historically recommended resource is other released funds, the sum of the number of clicks on fund A in the preset time period and the average number of clicks on other released funds in the preset time period may be calculated as the comprehensive number of clicks, the sum of the exposure of fund A in the preset time period and the average exposure of other released funds in the preset time period may be calculated as the comprehensive exposure, the ratio of the comprehensive number of clicks to the comprehensive exposure may be calculated to obtain the comprehensive click rate, the comprehensive click rate may be used as the target interaction factor, and the comprehensive exposure may be used as the target display factor. The resource to be recommended is the published news B, and the historical recommended resources are other published news of the same category as news B. The sum of the number of comments on news B within the preset time period and the average number of comments on other published news of the same category as news B within the preset time period can be calculated as the comprehensive number of comments. The sum of the exposure of fund A within the preset time period and the average exposure of other published funds within the preset time period can be calculated as the comprehensive exposure. The ratio of the comprehensive number of comments and the comprehensive exposure is calculated to obtain the comprehensive comment rate. The comprehensive comment rate is used as the target interaction factor, and the comprehensive exposure is used as the target display factor.

[0070] Step S208: Perform sample smoothing processing according to the target interaction factor and the target display factor to obtain the target smoothed interaction factor corresponding to the resource to be recommended.

[0071] Among them, since the interaction factor is calculated based on the interaction information and the display information, when any one of the interaction information and the display information changes, it will cause a large fluctuation and change in the interaction factor. Therefore, it is necessary to perform sample smoothing on the interaction factor to convert the interaction factor into a fast-iteration and relatively stable smooth interaction factor. Sample smoothing refers to smoothing the interaction factor data, filtering the noise of the interaction factor, and obtaining the smoothed interaction factor, that is, the smoothed interaction factor. Compared with the interaction factor, the smoothed interaction factor can better characterize the characteristics of the resource to be recommended, and the smoothed interaction factor is more credible and reliable. In one embodiment, the smoothed interaction factor is between 0 and 1. The larger the smoothed interaction factor, the higher the degree of interaction between the resource to be recommended and the public. For example, the interaction factor is the click-through rate of the resource to be recommended. After the click-through rate is sample smoothed, the click score corresponding to the resource to be recommended is obtained. The click score is between 0 and 1. The higher the click score, the more interested the public is in the resource to be recommended. The target smoothed interaction factor refers to the smoothed interaction factor corresponding to the resource to be recommended.

[0072] Specifically, the computer device may perform sample smoothing according to the target interaction factor and the target display factor to obtain the target smoothed interaction factor corresponding to the resource to be recommended. The computer device may specifically obtain a preset smoothing factor, scale the target display factor according to the preset smoothing factor, and fuse the target interaction factor according to the target display factor before and after scaling to obtain the target smoothed interaction factor.

[0073] Step S210, acquiring attribute information of the target user, performing information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtaining the matching degree between the resource to be recommended and the target user.

[0074] The target user refers to the user to whom resource recommendation is to be made. The user's attribute information may include at least one of the user's gender, age, education, region, hobbies, etc. The attribute information may characterize the user's characteristics to a certain extent. The matching degree may be a specific matching value. The larger the matching value, the higher the matching degree between the resource to be recommended and the target user. The matching degree may also be a matching probability. The larger the matching probability, the higher the matching degree. The matching degree may also be a label that characterizes whether the matching is successful. For example, 0 indicates a failed matching and 1 indicates a successful matching.

[0075] Specifically, the computer device may obtain attribute information of the target user, perform information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtain the matching degree between the resource to be recommended and the target user.

[0076] In one embodiment, information matching is performed based on attribute information and a target smooth interaction factor corresponding to the resource to be recommended to obtain a matching degree between the resource to be recommended and the target user, including: obtaining a trained information matching model; inputting the attribute information and the target smooth interaction factor into the trained information matching model to obtain a matching degree between the resource to be recommended and the target user.

[0077] The information matching model can be the result of learning one or more algorithms in machine learning and deep learning based on training data, and the information matching model is used to predict the matching degree between the recommended resources and the user. The algorithm can be support vector machine (SVM), adaptive boost algorithm (Adaptive Boost), convolutional neural network (Constitutional Neural Networks, CNN), etc.

[0078] Specifically, the computer device can obtain a trained information matching model, input the attribute information of the target user and the target smooth interaction factor corresponding to the resource to be recommended into the information matching model, and the information matching model outputs the matching degree between the resource to be recommended and the target user.

[0079] In one embodiment, there can be multiple target smoothed interaction factors input into the information matching model. For example, the click rate of the recommended resource is sample smoothed to obtain a click score, the subscription rate of the recommended resource is sample smoothed to obtain a subscription score, and the comment rate of the recommended resource is sample smoothed to obtain a comment score. The click score, subscription score, comment score of the recommended resource and the attribute information of the target user are input into the information matching model together to obtain the matching degree between the recommended resource and the target user.

[0080] In one embodiment, the generation of an information matching model includes the following steps: obtaining training samples of the information matching model to be trained; the training samples include a training smooth interaction factor corresponding to the training recommendation resource, attribute information of the training user, and matching labels of the training recommendation resource and the training user; using the training smooth interaction factor corresponding to the training recommendation resource and the attribute information of the training user as inputs to the information matching model to obtain predicted matching results of the training recommendation resource and the training user; adjusting the model parameters of the information matching model based on the predicted matching results and the matching labels until the convergence conditions are met to obtain a trained information matching model.

[0081] Among them, the training recommendation resource refers to the resource that has been published for training information matching model. The training smooth interaction factor refers to the smooth interaction factor corresponding to the training recommendation resource. It can be understood that the training recommendation resource can obtain the corresponding training smooth interaction factor based on the same processing as the resource to be recommended. The matching label of the training recommendation resource and the training user can be a specific matching value, a matching probability, or whether the matching is successful. The training sample includes multiple groups of training data, each group of training data includes the smooth interaction factor corresponding to the training recommendation resource, the attribute information of the training user, and the matching label of the training recommendation resource and the training user. The training recommendation resources in different groups of training data can be the same or different, and the training users in different groups of training data can be the same or different. For example, a group of training data includes the smooth interaction factor corresponding to fund A, the attribute information of user 1, and the matching label of fund A and user 1. A group of training data includes the smooth interaction factor corresponding to fund A, the attribute information of user 2, and the matching label of fund A and user 2. A group of training data includes the smooth interaction factor corresponding to fund B, the attribute information of user 2, and the matching label of fund B and user 2. A set of training data includes the smooth interaction factor corresponding to fund C, the attribute information of user 3, and the matching labels of fund C and user 3.

[0082] Specifically, in the information matching model to be trained, the input variables and output variables of the information matching model are configured in advance, the smooth interaction factor corresponding to the recommended resource and the attribute information of the user are set as the input variables of the information matching model, and the matching degree of the recommended resource and the user is set as the output variable of the information matching model. The computer device can obtain the training smooth interaction factor corresponding to the training recommended resource and the attribute information of the training user in a set of training data as the input of the information matching model, and use the matching labels of the training recommended resource and the training user in the set of training data as the expected output, and perform supervised training on the information matching model. The computer device trains the information matching model according to multiple sets of training data. During the training process, the computer device performs back propagation based on the predicted matching results and matching labels of the training recommended resource and the training user, adjusts the model parameters of the information matching model and continues training until the convergence condition is met, then the training is completed, and the trained information matching model is obtained.

[0083] In one embodiment, the computer device can train multiple information matching models based on different training samples, and different information matching models correspond to different resource topics, thereby improving the adaptability and accuracy of the information matching models. For example, an information matching model for predicting the matching degree between a fund to be recommended and a target user is obtained by training based on training samples related to funds. An information matching model for predicting the matching degree between a video to be recommended and a target user is obtained by training based on training samples related to videos. An information matching model for predicting the matching degree between news information to be recommended and a target user is obtained by training based on training samples related to news information.

[0084] In one embodiment, the resource attribute information of the training recommended resource, the smooth interaction factor and the attribute information of the training user can also be used as the input of the information matching model to train the information matching model and improve the accuracy of the information matching model. The resource attribute information can include at least one of the resource theme, resource category, resource keyword, resource data volume and resource publisher information. Accordingly, when the information matching model is used, the resource attribute information of the resource to be recommended, the target smooth interaction factor and the attribute information of the target user are input into the information matching model to obtain the matching degree between the resource to be recommended and the target user.

[0085] In traditional technology, in the recommendation scenario, deep learning models usually encounter a small or insufficient sample size during training, which is related to the magnitude of user items in the recommendation scenario and the particularity of the collected samples. For the extremely important indicator feature information (i.e., target interaction factor) in the recommendation scenario, such as click-through rate, comment rate, order rate, forwarding rate, etc., when the sample size is very small or insufficient, for example, when the exposure is small, the fluctuation of its calculated value will become very large. Directly using the indicator feature information without sample smoothing for model training will introduce large noise and affect the fitting of the model. In the above embodiment, the indicator feature information of the training recommendation resource is sample smoothed to obtain the training smooth interaction factor corresponding to the training recommendation resource, and the training smooth interaction factor is applied to the training of the information matching model, which can improve the effectiveness of the model training. Subsequently, the trained information matching model can obtain an effective and more accurate matching degree between the recommended resource and the target user, and effective resource recommendations can be made to the target user based on the matching degree.

[0086] Step S212: recommending resources to be recommended according to the matching degree.

[0087] Specifically, a matching degree threshold can be pre-set. When the matching degree between the resource to be recommended and the target user is greater than the matching degree threshold, the resource to be recommended is recommended to the target user. When there are multiple resources to be recommended, the computer device can obtain the matching degree between each resource to be recommended and the target user, sort each resource to be recommended according to the order of matching degree from large to small, and recommend a preset number of resources to be recommended with a high ranking to the target user, or determine the recommendation order of each resource to be recommended to the target user according to the sorting order of each resource to be recommended. The computer device can also assign different display parameters to each resource to be recommended according to the size of the matching degree, and display each resource to be recommended accordingly according to the display parameters. The display parameters determine the display form of the resource to be recommended, including display position, display size, dynamic display or static display, display fuzziness, display color, etc. For example, the higher the matching score, the closer the display position is and the clearer the display is. The resource to be recommended with a higher matching score can be set with animation effects.

[0088] In the above resource recommendation method, the target interaction information and target display information of the resource to be recommended are obtained; the historical interaction information and historical display information of the historically recommended resource are obtained; the target interaction factor and the target display factor are determined based on the target interaction information, the target display information, the historical interaction information and the historical display information; sample smoothing is performed according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended; the attribute information of the target user is obtained, and information matching is performed based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended to obtain the matching degree between the resource to be recommended and the target user; and the resource to be recommended is recommended according to the matching degree. In this way, the calculation of the target smooth interaction factor of the resource to be recommended comprehensively considers its own interaction information and display information as well as the interaction information and display information of the historically recommended resource, so that the target smooth interaction factor can better reflect the characteristics of the resource to be recommended. The resource to be recommended is recommended to the target user according to the matching degree, and the calculation of the matching degree comprehensively considers the attribute information of the target user and the target smooth interaction factor of the resource to be recommended, so that the recommendation of the resource is more in line with the characteristics of the user, the recommendation is more accurate, and the effectiveness of the resource recommendation is improved. Effective resource recommendations can avoid users from repeatedly searching or refreshing the interface due to invalid resource recommendations. Repeated searches or repeated refreshes of the interface will take up a lot of computer device resources. Therefore, on the basis of improving the effectiveness of resource recommendations, it can also reduce the waste of terminal or server resources.

[0089] In one embodiment, before obtaining the target interaction information and target display information of the resource to be recommended, the method also includes: obtaining the resource category of the resource to be recommended; obtaining reference interaction information and reference display information corresponding to the resource category; using the reference interaction information as the target interaction information, and using the reference display information as the target display information.

[0090] The reference interaction information is determined based on the statistical value of the interaction information of multiple published resources under the same resource category within a preset time period. The reference display information is determined based on the statistical value of the display information of multiple published resources under the same resource category within a preset time period. For example, the average value of the interaction information of multiple published resources under a resource category within a preset time period can be calculated, and the average value is used as the reference interaction information. The average value of the display information of multiple published resources under the same resource category within a preset time period can be calculated, and the average value is used as the reference display information.

[0091] Specifically, when the resource to be recommended is a newly released resource, the interactive information and display information of the resource to be recommended itself are zero, and the interactive information and display information will make it basically impossible to recommend the newly released resource to the user. At this time, the computer device can obtain the resource category of the resource to be recommended, obtain the reference interactive information corresponding to the resource category, use the reference interactive information as the target interactive information, obtain the reference display information corresponding to the resource category, and use the reference display information as the target display information. For example, the newly released fund to be recommended is a 5G fund, which belongs to the technology fund. Then the average of the clicks and exposures of all or part of the funds under the technology fund can be calculated, and the average of the clicks can be used as the clicks of the fund to be recommended, and the average of the exposure can be used as the exposure of the fund to be recommended.

[0092] In this embodiment, when the resource to be recommended is a newly released resource, the reference interaction information and reference display information corresponding to the resource category of the resource to be recommended can be obtained as the target interaction information and target display information. In this way, the situation where the target interaction information and target display information are zero, which affects the resource recommendation, can be avoided, thereby improving the effectiveness of resource recommendation.

[0093] In one embodiment, before obtaining the target interaction information and target display information of the resource to be recommended, the method also includes: obtaining resource attribute information of the resource to be recommended; inputting the resource attribute information into a trained interaction factor prediction model to obtain a reference interaction factor; the interaction factor prediction model is trained based on the resource attribute information and the training interaction factor of the training recommended resource; obtaining the resource category of the resource to be recommended, obtaining the reference display information corresponding to the resource category, and using the reference display information as the target display information; and obtaining the target interaction information based on the reference interaction factor and the reference display information.

[0094] Among them, resource attribute information may include at least one of resource theme, resource category, resource keyword, resource data volume and resource publisher information. Resource keywords can be extracted from the recommended resources using various keyword extraction algorithms, such as TF-IDF algorithm (Term Frequency-Inverse Document Frequency), Word2Vec algorithm, TextRank algorithm, etc. The title and tag of the recommended resource can also be used as resource keywords. The resource data volume can be article length, image size, video length, audio length, etc. Resource publisher information refers to the information of the user who publishes the resource, such as the author of the news and the publisher of the news.

[0095] The interaction factor prediction model can be the result of learning one or more algorithms in machine learning and deep learning based on training data, and is used to predict the interaction factor of resources. The algorithm can be support vector machine (SVM), adaptive boost algorithm (Adaptive Boost), convolutional neural network (CNN), etc.

[0096] Specifically, the computer device can obtain a trained interaction factor prediction model, input the resource attribute information of the resource to be recommended into the interaction factor prediction model, and the interaction factor prediction model outputs a reference interaction factor of the resource to be recommended. Further, the computer device obtains the resource category of the resource to be recommended, obtains the reference display information corresponding to the resource category, uses the reference display information as the target display information, and obtains the target interaction information based on the reference interaction factor and the target display information. For example, the resource attribute information of the fund to be recommended is input into the fund click-through rate prediction model to obtain the click-through rate of the fund to be recommended. The average exposure corresponding to the fund category of the fund to be recommended is obtained as the reference exposure, and the reference exposure and the click-through rate predicted by the model are multiplied to obtain the click volume of the fund to be recommended.

[0097] In the interactive factor prediction model, the input variables and output variables of the interactive factor prediction model are configured in advance, the resource attribute information of the recommended resources is set as the input variable of the interactive factor prediction model, and the interactive factor of the recommended resources is set as the output variable of the interactive factor prediction model. The computer device can obtain the resource attribute information of the training recommended resources as the input of the interactive factor prediction model, and use the training interactive factor of the training recommended resources as the expected output to perform supervised training on the interactive factor prediction model. The computer device trains the interactive factor prediction model based on multiple sets of training data. During the training process, the computer device performs back propagation based on the prediction results of the training recommended resources and the training interactive factor, adjusts the model parameters of the interactive factor prediction model and continues training until the convergence condition is met, then the training is completed, and the trained interactive factor prediction model is obtained.

[0098] In one embodiment, the computer device can train interaction factor prediction models corresponding to different interaction factors of different resource topics, for example, a fund click rate prediction model for predicting fund click rate, a fund subscription rate prediction model for predicting fund subscription rate, and a news comment rate prediction model for predicting news information comment rate.

[0099] In the above embodiment, when the resource to be recommended is a newly released resource, the reference interaction factor can be predicted by the interaction factor prediction model, the reference display information corresponding to the resource category of the resource to be recommended can be obtained as the target display information, and the target interaction information can be determined according to the reference interaction factor and the reference display information. In this way, the situation where the target interaction information and the target display information are zero, which affects the resource recommendation, can be avoided, thereby improving the effectiveness of resource recommendation.

[0100] In one embodiment, Figure 3 As shown, the target interaction factor and the target display factor are determined based on the target interaction information, the target display information, the historical interaction information and the historical display information, including:

[0101] Step S302: Obtain interaction fusion information according to the target interaction information and the historical interaction information.

[0102] Step S304: Obtain display fusion information according to the target display information and the historical display information.

[0103] Step S306: taking the ratio of the interactive fusion information to the display fusion information as the target interactive factor, and taking the display fusion information as the target display factor.

[0104] Specifically, the computer device may obtain interactive fusion information according to the target interactive information and the historical interactive information, obtain display fusion information according to the target display information and the historical display information, use the ratio of the interactive fusion information to the display fusion information as the target interactive factor, and use the display fusion information as the target display factor. Specifically, the sum or weighted sum of the target interactive information and the historical interactive information may be used as the interactive fusion information, and the sum or weighted sum of the target display information and the historical display information may be used as the display fusion information.

[0105] For example, the number of clicks on the resource to be recommended is click, the exposure of the resource to be recommended is impression, and the average number of clicks on multiple historical recommended resources belonging to the same resource category as the resource to be recommended is click. b The target interaction factor is The target display factor is impression+impression b .

[0106] In one embodiment, Figure 3 As shown, sample smoothing is performed according to the target interaction factor and the target display factor to obtain the target smoothed interaction factor corresponding to the resource to be recommended, including:

[0107] Step S308: Obtain a first preset smoothing factor.

[0108] Step S310: scaling the target presentation factor according to a first preset smoothing factor to obtain a target scaled presentation factor.

[0109] Step S312: fusing the target interaction factor according to the target scaling display factor and the target display factor to obtain a target smooth interaction factor.

[0110] Specifically, the computer device can obtain a first preset smoothing factor, scale the target display factor according to the first preset smoothing factor to obtain a target scaled display factor, and fuse the target interaction factor according to the target scaled display factor and the target display factor to obtain a target smoothed interaction factor.

[0111] In one embodiment, the smooth interaction factor can be calculated by the following formula:

[0112]

[0113] Among them, wilson(p,n) represents the target smooth interaction factor, p represents the target interaction factor, n represents the target display factor, and z represents the preset smoothing factor. Both represent scaling factors. z is the statistic corresponding to a certain confidence level of the Wilson confidence interval, which can be obtained by looking up the table. Generally, at a 95% confidence level, the value of z is 1.96.

[0114] For example, the number of clicks on the resource to be recommended is click, the exposure of the resource to be recommended is impression, and the average number of clicks on multiple historical recommended resources belonging to the same resource category as the resource to be recommended is click. b , the average exposure is impression b The click score of the resource to be recommended (i.e., the target smooth interaction factor) is

[0115]

[0116] In this embodiment, a comprehensive target interaction factor and a target display factor are obtained by combining the target interaction information and target display information of the resource to be recommended and the historical interaction information and historical display information of the historically recommended resource, and a target smoothed interaction factor is obtained by performing sample smoothing processing based on the target interaction factor and the target display factor. In this way, the problem that when the resource to be recommended is a newly released resource, the interaction information and display information of the newly released resource cannot be recommended to the user due to the large gap between the interaction information and display information of the released resource can be solved. For example, the click rate and exposure of the new fund are 0, and the average click rate of the fund market is 6%. If the click score of the new fund is directly calculated based on the click rate and exposure as 0, then the new fund will be due to the large gap with the existing funds due to the click score of 0, which will cause the new fund to be basically unable to be recommended to the user. However, the click score of the new fund calculated according to the method of this embodiment is not 0, and the new fund can be recommended to the user later. In addition, if the click rates of Fund 1 and Fund 2 are the same, but the exposure of Fund 1 is greater than that of Fund 2, the click score of Fund 1 calculated according to the method of this embodiment is higher than the click score of Fund 2, which improves the credibility of the target smoothed interaction factor.

[0117] In one embodiment, Figure 4 As shown, the target interaction factor includes a first interaction factor and a second interaction factor, and the target display factor includes a first display factor and a second display factor. The target interaction factor and the target display factor are determined based on the target interaction information, the target display information, the historical interaction information, and the historical display information, including:

[0118] Step S402: taking the ratio of the target interaction information to the target display information as the first interaction factor, and taking the target display information as the first display factor.

[0119] Step S404: taking the ratio of the historical interaction information to the historical display information as the second interaction factor, and taking the historical display information as the second display factor.

[0120] Specifically, the target interaction factor includes a first interaction factor and a second interaction factor, and the target display factor includes a first display factor and a second display factor. The computer device can calculate the ratio of the target interaction information and the target display information to obtain the first interaction factor, use the target display information as the first display factor, calculate the ratio of the historical interaction information and the historical display information as the second interaction factor, and use the historical display information as the second display factor.

[0121] For example, the number of clicks on the resource to be recommended is click, the exposure of the resource to be recommended is impression, and the average number of clicks on multiple historical recommended resources belonging to the same resource category as the resource to be recommended is click. b , the average exposure is impression b The first interaction factor is The second interaction factor is The first display factor is impression, and the second display factor is impression b .

[0122] In one embodiment, Figure 4 As shown, sample smoothing is performed according to the target interaction factor and the target display factor to obtain the target smoothed interaction factor corresponding to the resource to be recommended, including:

[0123] Step S406, obtaining a second preset smoothing factor.

[0124] Step S408: scaling the first presentation factor according to the second preset smoothing factor to obtain a first scaled presentation factor.

[0125] Step S410: The first interaction factor is fused according to the first scaling display factor and the first display factor to obtain a first smoothing interaction factor.

[0126] Specifically, the computer device can obtain a second preset smoothing factor, scale the first display factor according to the second preset smoothing factor to obtain a first scaled display factor, and fuse the first interaction factor according to the first scaled display factor and the first display factor to obtain a first smoothed interaction factor.

[0127] For example, the number of clicks on the recommended resource is click, the exposure of the recommended resource is impression, and the first smooth interaction factor is The first zoom display factor includes Where n is impression.

[0128] Step S412: scaling the second presentation factor according to the second preset smoothing factor to obtain a second scaled presentation factor.

[0129] Step S414: The second interaction factor is fused according to the second scaling display factor and the second display factor to obtain a second smooth interaction factor.

[0130] Specifically, the computer device may scale the second display factor according to the second preset smoothing factor to obtain a second scaled display factor, and fuse the second interaction factor according to the second scaled display factor and the second display factor to obtain a second smoothed interaction factor.

[0131] For example, the average number of clicks for multiple historical recommended resources that belong to the same resource category as the resource to be recommended is click b , the average exposure is impression b , the second smooth interaction factor is The second zoom display factor includes Where n is impression b .

[0132] Step S416, obtaining a target smooth interaction factor according to the first smooth interaction factor and the second smooth interaction factor.

[0133] Specifically, after the first smooth interaction factor and the second smooth interaction factor are calculated, the computer device may obtain the target smooth interaction factor according to the first smooth interaction factor and the second smooth interaction factor. Specifically, the target smooth interaction factor may be obtained by calculating the sum or weighted sum of the first smooth interaction factor and the second smooth interaction factor.

[0134] In one embodiment, a target smooth interaction factor is obtained according to a first smooth interaction factor and a second smooth interaction factor, including: obtaining a first proportional adjustment factor corresponding to the first smooth interaction factor, obtaining a second proportional adjustment factor corresponding to the second smooth interaction factor; adjusting the first smooth interaction factor and the second smooth interaction factor according to the first proportional adjustment factor and the second proportional adjustment factor to obtain the target smooth interaction factor.

[0135] Specifically, the computer device can obtain the weighting factors corresponding to the first smooth interaction factor and the second smooth interaction factor (ie, the first proportional adjustment factor and the second proportional adjustment factor), and calculate the weighted sum of the first smooth interaction factor and the second smooth interaction factor based on the weighting factors to obtain the target smooth interaction factor.

[0136] In one embodiment, the target smooth interaction factor can be calculated by the following formula:

[0137]

[0138]

[0139] Among them, y represents the target smooth interaction factor, and factor represents the weighting factor.

[0140] In the above embodiment, the target interaction information and target display information of the resource to be recommended are first considered separately to obtain the first interaction factor and the first display factor, the first smoothed interaction factor is obtained by performing sample smoothing processing based on the first interaction factor and the first display factor, the second interaction factor and the second display factor are considered separately to obtain the second interaction factor and the second display factor, the second smoothed interaction factor is obtained by performing sample smoothing processing based on the second interaction factor and the second display factor, and finally the target smoothed interaction factor is obtained by comprehensively considering the first smoothed interaction factor and the second smoothed interaction factor. In this way, the problem caused by comprehensively considering the target interaction information and the target display information of the resource to be recommended and the historical interaction information and the historical display information of the historically recommended resource at one time can be solved, and the accuracy of the target smoothed interaction factor can be improved.

[0141] For example, assuming that the average number of clicks is 14, the average exposure is 200, and the average click rate is 7%, the comparison of click scores of resource 1 and resource 2 is shown in Table 1.

[0142] Table 1

[0143]

[0144] As shown in Table 1, the click rate and click score calculated by Formula 1 of Resource 1 are better than those of Resource 2, but the click score calculated by Formula 2 of Resource 1 is lower than that of Resource 2. This shows that Formula 2 may have problems in a few cases. It can be seen that when the exposure of resource 1 is low and the click rate of resource 1 is significantly higher than the average click rate, the formula If the influence of the numerator is greater than that of the denominator, then even when the click-through rate of resource 1 is significantly better than that of resource 2, the click score of resource 1 calculated by formula 2 will be lower than the click score of resource 2 calculated by formula 2. Similarly, when the exposure of resource 1 is low and the click-through rate of resource 1 is significantly lower than the average click-through rate, the above problem will also occur. In summary, when the interaction factor of the resource to be recommended is significantly lower than or higher than the interaction factor of the historically recommended resource, the target smoothed interaction factor obtained by sample smoothing the data of the resource to be recommended by formula 2 is inaccurate. However, the method according to this embodiment (i.e., formula 3) can solve the above problem and improve the accuracy of the target smoothed interaction factor.

[0145] In one embodiment, obtaining a first proportional adjustment factor corresponding to a first smooth interaction factor and obtaining a second proportional adjustment factor corresponding to a second smooth interaction factor include: obtaining a confidence interval and expectation corresponding to the first smooth interaction factor; obtaining the first proportional adjustment factor according to the distance between the confidence interval and the target expectation; and determining the second proportional adjustment factor based on the first proportional adjustment factor.

[0146] Specifically, refer to Figure 5 , Figure 5 The upper and lower bounds of the confidence interval corresponding to Formula 1 and the schematic diagram of the expected change curve are shown in Figure 1. The horizontal axis represents the exposure and the vertical axis represents the smooth interaction factor. Figure 5 As shown, the confidence of the display information and the interactive information, with reference to the distance between the upper and lower bounds and the expectation, is not a linear relationship. This nonlinear change in which the upper and lower bounds gradually approach the expectation as the display information increases is introduced into the construction of the weighting factor. The computer device can obtain the confidence interval and expectation corresponding to the first smooth interaction factor. The confidence interval includes an upper bound and a lower bound, and the relative distance between the upper and lower bounds and the expectation (that is, the ratio of the sum of the distance between the upper and lower bounds and the expectation to the expectation) is used as the first proportional adjustment factor. The sum of the first proportional adjustment factor and the second proportional adjustment factor is 1. After determining the first proportional adjustment factor, the difference between 1 and the first proportional adjustment factor can be used as the second proportional adjustment factor.

[0147] In one embodiment, the smooth interaction factor is The upper bound of the confidence interval is The lower bound of the confidence interval is Expected to be According to the relative distance between the upper and lower bounds of the confidence interval and the expectation, it can be calculated Wherein, n represents the display information of the resource to be recommended, and p represents the interaction factor of the resource to be recommended.

[0148] In this embodiment, the first proportional adjustment factor and the second proportional adjustment factor can be quickly obtained according to the confidence interval and expectation corresponding to the first smooth interaction factor.

[0149] In one embodiment, obtaining a first proportional adjustment factor corresponding to a first smooth interaction factor and obtaining a second proportional adjustment factor corresponding to a second smooth interaction factor include: obtaining a confidence interval corresponding to the first smooth interaction factor; calculating the width of the confidence interval to obtain the first proportional adjustment factor; and determining the second proportional adjustment factor based on the first proportional adjustment factor.

[0150] Specifically, refer to Figure 5, as the displayed information increases, the width of the confidence interval will gradually decrease, so the computer device can obtain the confidence interval corresponding to the first smooth interaction factor, and use the width of the confidence interval as the first proportional adjustment factor. The sum of the first proportional adjustment factor and the second proportional adjustment factor is 1, so after determining the first proportional adjustment factor, the difference between 1 and the first proportional adjustment factor can be used as the second proportional adjustment factor.

[0151] In one embodiment, the smooth interaction factor is The upper bound of the confidence interval is The lower bound of the confidence interval is The distance between the upper and lower bounds of the confidence interval can be calculated Wherein, n represents the display information of the resource to be recommended, and p represents the interaction factor of the resource to be recommended.

[0152] In this embodiment, the first proportional adjustment factor and the second proportional adjustment factor can be quickly obtained according to the confidence interval corresponding to the first smooth interaction factor. In addition, the change of factor1 mainly depends on the size growth of np. For the recommended information with a smaller interaction factor, the change speed of factor1 is too slow. For example, when the click-through rate and exposure of the recommended resource are 0, factor1 = 1. When the click-through rate of the recommended resource is 1% and the exposure is 100, factor1≈0.87. However, when p is far away from 0.5, under the same n, the width of the confidence interval will decrease faster, that is, as the interaction factor decreases or increases, the change rate of factor2 will accelerate.

[0153] In one embodiment, obtaining a first proportional adjustment factor corresponding to a first smooth interaction factor and obtaining a second proportional adjustment factor corresponding to a second smooth interaction factor include: obtaining an interactive influence factor corresponding to the first interactive factor; adjusting the first proportional adjustment factor according to the interactive influence factor; and determining the second proportional adjustment factor based on the adjusted first proportional adjustment factor.

[0154] Specifically, if it is desired that the change speed of the weighted factor is equal under any interaction factor, the computer device can obtain the interaction influence factor corresponding to the first interaction factor, adjust and correct the first proportional adjustment factor according to the interaction influence factor, and obtain the adjusted first proportional adjustment factor. The sum of the first proportional adjustment factor and the second proportional adjustment factor is 1, so when the adjusted first proportional adjustment factor is determined, the difference between 1 and the adjusted first proportional adjustment factor can be used as the second proportional adjustment factor.

[0155] In one embodiment, In order to make the speed at which factor changes with the change of n tend to be stable, the influence of p can be eliminated by simplifying p(1-p). Specifically, the maximum value of p(1-p) can be obtained to replace p(1-p), and we can get refer to Figure 6 , the weighting factor factor3 gradually decreases as n increases, but the speed of change will gradually slow down, which is in line with the expected characteristics of the weighting factor that adapts to the displayed information. Because as n increases, the confidence interval will gradually shrink, and the shrinking speed itself will gradually slow down. For example, for the same click-through rate, the change in the confidence interval and credibility is not equal when the exposure increases from 100 to 1000, but gradually slows down when it increases from 10000 to 10900. Therefore, the gradual slowdown of the change speed of the weighting factor factor3 is in line with expectations, and when z>0, the value of factor3 is between 0 and 1, which meets the requirements of the weighting factor in the weighted calculation. In this way, the target smoothed interaction factor obtained by sample smoothing through formula 3+factor3 can be more accurate, reliable, and credible.

[0156] The present application also provides an application scenario, which applies the above-mentioned resource recommendation method, and the method is applied to a financial recommendation scenario. Specifically, taking the resource as a fund as an example, the application of the resource recommendation method in the application scenario is as follows:

[0157] 1. Determine the click volume, order volume and impression volume of the fund to be recommended in the past month, and determine the average click volume of multiple historical recommended funds belonging to the same fund category as the fund to be recommended in the past month. b , average order quantity b and average impression b .

[0158] 2. According to click, impression, and average click b and average impression b Determine the click score of the fund to be recommended. b and average impression b Determine the subscription score of the fund to be recommended.

[0159] Specifically, it can be done by Perform sample smoothing and calculate the click score of the fund to be recommended.

[0160] Calculate the click score of the fund to be recommended. Factor can be It can also be It can also be When calculating factor, n is impression and p is Similarly, the subscription score of the fund to be recommended can be calculated using the above formula.

[0161] Furthermore, sample smoothing based on weighted factors (i.e., formula 3+factor3) can more effectively characterize the statistical characteristics of recommended resources in the case of small sample sizes. As shown in Table 2, the funds to be recommended include Fund 1, Fund 2, and Fund 3. The click scores of the funds to be recommended are calculated based on the click volume, exposure volume of the funds to be recommended and the click volume and exposure volume of the large-cap funds through sample smoothing based on weighted factors. As shown in Table 2, the click score of Fund 2 is higher than that of Fund 1 and Fund 3. On the one hand, this is because Fund 2 has a better exposure volume when the click rate is similar. On the other hand, as the exposure volume changes, different funds correspond to different weighted factors, which effectively balances the dynamic effects between the click volume, exposure volume of the funds to be recommended and the click volume and exposure volume of the historically recommended funds.

[0162] Table 2

[0163]

[0164] For further reference, Figure 7 , Figure 7 The figure is a schematic diagram of the comparison before and after the sample smoothing process. The horizontal axis represents the exposure, and the vertical axis represents the click-through rate or click score. The average click-through rate of large-cap funds is 10%, the average exposure is 100, and the original click-through rate of the recommended fund is 20%. Curve 1 represents the curve of the original click-through rate (i.e., before the sample smoothing process) changing with the change of exposure, Curve 2 represents the curve of the click score (i.e., after the sample smoothing process is performed by formula 3+factor3) changing with the change of exposure, and Curve 3 represents the target interaction factor (i.e., ) is a curve that changes with the exposure amount. According to curve 1, curve 2 and curve 3, it can be seen that the iterative convergence speed of curve 2 is faster than that of curve 3, and the iterative convergence speed of curve 2 is more stable than that of curve 1.

[0165] 3. Input the click score, order score, fund attribute information and target user attribute information of the fund to be recommended into the information matching model to obtain the matching degree between the fund to be recommended and the target user.

[0166] 4. Recommend the recommended funds to the target users based on the matching degree.

[0167] The terminal or server can periodically obtain the interaction-related information and display-related information of each fund to be recommended in the candidate fund pool to calculate the click score and subscription score of the fund to be recommended. The target user can click on the financial application on the terminal and enter the fund recommendation interface after completing the login operation. The terminal or server can obtain attribute information from the target user's login information, obtain the fund attribute information of the fund to be recommended from the candidate fund pool, obtain the calculated click score and subscription score of the fund to be recommended, and input the click score, subscription score, fund attribute information of the fund to be recommended and the attribute information of the target user into the information matching model to obtain the matching degree of the fund to be recommended and the target user. Finally, the terminal determines the display order of each fund to be recommended based on the matching degree of each fund to be recommended and the target user, and displays the fund to be recommended to the target user on the fund recommendation interface according to the display order. Figure 8 As shown, according to the matching degree between the recommended fund and the target user, the recommended order of the new energy vehicle fund is determined to be the first, the recommended order of the navigation equipment fund is the second, and the recommended order of the communication equipment fund is the third. It can be understood that the higher the matching degree and the higher the recommended order, the more interested the user will be.

[0168] The present application also provides an application scenario, which applies the above resource recommendation method, and the method is applied to a news recommendation scenario. Specifically, taking news as an example, the resource recommendation method is applied to the application scenario as follows:

[0169] 1. Determine the click, comment, send and impression of the news to be recommended in the past three months, and determine the average click of multiple news in the same news category as the news to be recommended in the past three months. b , average number of comments b , average forwarding volume send b and average impression b .

[0170] 2. According to click, impression, and average click b and average impression b Determine the click score of the news to be recommended. Based on the number of comments, impression, and average number of comments. b and average impression b Determine the comment score of the news to be recommended. Based on the forwarding volume send, impression volume, and average forwarding volume send b and average impression bDetermine the forwarding score of the news to be recommended.

[0171] 3. Input the click score, comment score, forwarding score, news attribute information and target user attribute information of the news to be recommended into the information matching model to obtain the matching degree between the news to be recommended and the target user.

[0172] 4. Recommend news to target users based on the matching degree.

[0173] The server can periodically obtain the interaction-related information and display-related information of each news to be recommended in the candidate news pool to calculate the click score, comment score, and forwarding score of the news to be recommended. The target user can click on a social application, such as Weibo, on the terminal to enter the news recommendation interface after completing the login operation. When it is detected that the target user has entered the news recommendation interface, the terminal sends a news recommendation request to the server, and the news recommendation request carries the user ID of the target user. The server can obtain the login information of the target user based on the user ID, obtain the attribute information from the login information of the target user, obtain the news attribute information of the news to be recommended from the candidate news pool, obtain the click score, comment score, and forwarding score of the news to be recommended, and input the click score, comment score, forwarding score, news attribute information of the news to be recommended and the attribute information of the target user into the information matching model to obtain the matching degree of the news to be recommended and the target user. The server returns the matching degree of each news to be recommended and the target user to the terminal. Finally, the terminal determines the display order of each news to be recommended based on the received matching degree, and displays the news to be recommended to the target user on the news recommendation interface according to the display order. Fig. 9 As shown, according to the matching degree between the news to be recommended and the target user, the news published by user A is recommended first, the news published by user B is recommended second, and the news published by user C is recommended third. It can be understood that the higher the matching degree and the higher the recommendation order, the more interested the user will be.

[0174] In the above technical scheme, data smoothing is performed on the relevant information of the fund (or news) to be recommended in combination with the relevant information of the historically recommended funds (or news), so that data that can better reflect the characteristics of the fund (or news) to be recommended can be constructed. Based on the data, more effective fund (or news) recommendations can be made and the waste of resources of the terminal or server can be reduced.

[0175] It should be understood that although Figure 2-4 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-4At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0176] In one embodiment, Fig.10 As shown, a resource recommendation device is provided. The device can adopt a software module or a hardware module, or a combination of the two to become a part of a computer device. The device specifically includes: an information acquisition module 1002, a factor determination module 1004, an information smoothing module 1006, a matching degree determination module 1008 and a resource recommendation module 1010, wherein:

[0177] Information acquisition module 1002, used to acquire target interaction information and target display information of the resource to be recommended; acquire historical interaction information and historical display information of the historically recommended resource;

[0178] A factor determination module 1004, configured to determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information;

[0179] An information smoothing module 1006 is used to perform sample smoothing processing according to the target interaction factor and the target display factor to obtain a target smooth interaction factor corresponding to the resource to be recommended;

[0180] A matching degree determination module 1008 is used to obtain attribute information of a target user, and perform information matching based on the attribute information and a target smooth interaction factor corresponding to the resource to be recommended, to obtain a matching degree between the resource to be recommended and the target user;

[0181] The resource recommendation module 1010 is used to recommend the resource to be recommended according to the matching degree.

[0182] In one embodiment, the information acquisition module is further used to acquire resource categories of resources to be recommended; acquire reference interaction information and reference display information corresponding to the resource categories; use the reference interaction information as target interaction information, and use the reference display information as target display information.

[0183] In one embodiment, the information acquisition module is also used to obtain resource attribute information of the resources to be recommended; input the resource attribute information into a trained interaction factor prediction model to obtain a reference interaction factor; the interaction factor prediction model is trained based on the resource attribute information of the training recommended resources and the training interaction factor; obtain the resource category of the resources to be recommended, obtain the reference display information corresponding to the resource category, and use the reference display information as the target display information; obtain the target interaction information based on the reference interaction factor and the reference display information.

[0184] In one embodiment, the factor determination module is also used to obtain interactive fusion information based on the target interactive information and the historical interactive information; obtain display fusion information based on the target display information and the historical display information; use the ratio of the interactive fusion information to the display fusion information as the target interactive factor, and use the display fusion information as the target display factor.

[0185] In one embodiment, the information smoothing module is also used to obtain a first preset smoothing factor; scale the target display factor according to the first preset smoothing factor to obtain a target scaled display factor; and fuse the target interaction factor according to the target scaled display factor and the target display factor to obtain a target smoothed interaction factor.

[0186] In one embodiment, the target interaction factor includes a first interaction factor and a second interaction factor, the target display factor includes a first display factor and a second display factor, and the factor determination module is further used to use the ratio of the target interaction information to the target display information as the first interaction factor, and the target display information as the first display factor; use the ratio of the historical interaction information to the historical display information as the second interaction factor, and use the historical display information as the second display factor.

[0187] In one embodiment, the information smoothing module is also used to obtain a second preset smoothing factor; scale the first display factor according to the second preset smoothing factor to obtain a first scaled display factor; fuse the first interaction factor according to the first scaled display factor and the first display factor to obtain a first smoothed interaction factor; scale the second display factor according to the second preset smoothing factor to obtain a second scaled display factor; fuse the second interaction factor according to the second scaled display factor and the second display factor to obtain a second smoothed interaction factor; obtain a target smoothed interaction factor according to the first smoothed interaction factor and the second smoothed interaction factor.

[0188] In one embodiment, the information smoothing module is also used to obtain a first proportional adjustment factor corresponding to the first smooth interaction factor, and to obtain a second proportional adjustment factor corresponding to the second smooth interaction factor; the first smooth interaction factor and the second smooth interaction factor are adjusted according to the first proportional adjustment factor and the second proportional adjustment factor to obtain a target smooth interaction factor.

[0189] In one embodiment, the information smoothing module is further used to obtain the confidence interval and expectation corresponding to the first smooth interaction factor; obtain the first proportional adjustment factor according to the distance between the confidence interval and the target expectation; and determine the second proportional adjustment factor based on the first proportional adjustment factor.

[0190] In one embodiment, the information smoothing module is further used to obtain a confidence interval corresponding to the first smooth interaction factor; calculate the width of the confidence interval to obtain a first proportional adjustment factor; and determine a second proportional adjustment factor based on the first proportional adjustment factor.

[0191] In one embodiment, the information smoothing module is further used to obtain an interaction influence factor corresponding to the first interaction factor; adjust the first proportional adjustment factor according to the interaction influence factor; and determine the second proportional adjustment factor based on the adjusted first proportional adjustment factor.

[0192] In one embodiment, the matching degree determination module is further used to obtain a trained information matching model; input the attribute information and the target smooth interaction factor into the trained information matching model to obtain the matching degree between the resource to be recommended and the target user.

[0193] In one embodiment, the information acquisition module is also used to obtain training samples of the information matching model to be trained; the training samples include the training smooth interaction factor corresponding to the training recommendation resource, the attribute information of the training user, and the matching labels of the training recommendation resource and the training user; the training smooth interaction factor corresponding to the training recommendation resource and the attribute information of the training user are used as inputs of the information matching model to obtain the predicted matching results of the training recommendation resource and the training user; the model parameters of the information matching model are adjusted based on the predicted matching results and the matching labels until the convergence conditions are met to obtain the trained information matching model.

[0194] For the specific definition of the resource recommendation device, please refer to the definition of the resource recommendation method above, which will not be repeated here. Each module in the above resource recommendation device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0195] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.11As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as interactive information, display information, information matching models, interactive factor prediction models, etc. of resources. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a resource recommendation method is implemented.

[0196] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.12 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a resource recommendation method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0197] Those skilled in the art will understand that Fig.11 , 12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0198] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0199] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0200] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.

[0201] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0202] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0203] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A resource recommendation method, characterized in that: The method comprises: Obtain target interaction information and target display information of the resource to be recommended; Obtain historical interaction information and historical display information of historical recommended resources; Determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information; Perform sample smoothing processing according to the target interaction factor and the target display factor to obtain the target smoothed interaction factor corresponding to the resource to be recommended; Acquire attribute information of the target user, perform information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended, and obtain the matching degree between the resource to be recommended and the target user; The resource to be recommended is recommended according to the matching degree.

2. The method according to claim 1, characterized in that Before obtaining the target interaction information and target display information of the resource to be recommended, the method further includes: Obtaining the resource category of the resource to be recommended; Obtaining reference interaction information and reference display information corresponding to the resource category; The reference interaction information is used as the target interaction information, and the reference display information is used as the target display information.

3. The method according to claim 1, characterized in that Before obtaining the target interaction information and target display information of the resource to be recommended, the method further includes: Obtaining resource attribute information of the resource to be recommended; Inputting the resource attribute information into a trained interaction factor prediction model to obtain a reference interaction factor; the interaction factor prediction model is trained based on the resource attribute information of the training recommended resources and the training interaction factor; Acquire the resource category of the resource to be recommended, acquire reference display information corresponding to the resource category, and use the reference display information as the target display information; The target interaction information is obtained according to the reference interaction factor and the reference display information.

4. The method according to claim 1, characterized in that: The determining of the target interaction factor and the target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information includes: Obtaining interactive fusion information according to the target interactive information and the historical interactive information; Obtaining display fusion information according to the target display information and the historical display information; The ratio of the interactive fusion information to the display fusion information is used as the target interactive factor, and the display fusion information is used as the target display factor.

5. The method according to claim 4, characterized in that The sample smoothing process is performed according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended, including: Obtaining a first preset smoothing factor; Scaling the target display factor according to the first preset smoothing factor to obtain a target scaled display factor; The target interaction factor is fused according to the target scaling display factor and the target display factor to obtain the target smooth interaction factor.

6. The method according to claim 1, characterized in that The target interaction factor includes a first interaction factor and a second interaction factor, the target display factor includes a first display factor and a second display factor, and determining the target interaction factor and the target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information includes: Using the ratio of the target interaction information to the target display information as the first interaction factor, and using the target display information as the first display factor; The ratio of the historical interaction information to the historical display information is used as the second interaction factor, and the historical display information is used as the second display factor.

7. The method according to claim 6, characterized in that The sample smoothing process is performed according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended, including: Obtaining a second preset smoothing factor; Scaling the first display factor according to the second preset smoothing factor to obtain a first scaled display factor; fusing the first interaction factor according to the first scaling display factor and the first display factor to obtain a first smoothing interaction factor; Scaling the second display factor according to the second preset smoothing factor to obtain a second scaled display factor; fusing the second interaction factor according to the second scaling display factor and the second display factor to obtain a second smoothing interaction factor; The target smooth interaction factor is obtained according to the first smooth interaction factor and the second smooth interaction factor.

8. The method according to claim 7, characterized in that The step of obtaining the target smooth interaction factor according to the first smooth interaction factor and the second smooth interaction factor includes: Obtain a first proportional adjustment factor corresponding to the first smooth interaction factor, and obtain a second proportional adjustment factor corresponding to the second smooth interaction factor; The first smooth interaction factor and the second smooth interaction factor are adjusted according to the first proportional adjustment factor and the second proportional adjustment factor to obtain the target smooth interaction factor.

9. The method according to claim 8, characterized in that The obtaining of a first proportional adjustment factor corresponding to the first smooth interaction factor and the obtaining of a second proportional adjustment factor corresponding to the second smooth interaction factor include: Obtaining the confidence interval and expectation corresponding to the first smooth interaction factor; Obtaining the first proportional adjustment factor according to the confidence interval and the target expected distance; The second scaling factor is determined based on the first scaling factor.

10. The method according to claim 8, characterized in that The obtaining of a first proportional adjustment factor corresponding to the first smooth interaction factor and the obtaining of a second proportional adjustment factor corresponding to the second smooth interaction factor include: Obtaining a confidence interval corresponding to the first smooth interaction factor; Calculating the width of the confidence interval to obtain the first proportional adjustment factor; The second scaling factor is determined based on the first scaling factor.

11. The method according to claim 10, characterized in that The obtaining of a first proportional adjustment factor corresponding to the first smooth interaction factor and the obtaining of a second proportional adjustment factor corresponding to the second smooth interaction factor include: Obtaining an interaction influence factor corresponding to the first interaction factor; Adjusting the first proportional adjustment factor according to the interactive influence factor; The second proportional adjustment factor is determined based on the adjusted first proportional adjustment factor.

12. The method according to claim 1, characterized in that The performing information matching based on the attribute information and the target smooth interaction factor corresponding to the resource to be recommended to obtain the matching degree between the resource to be recommended and the target user includes: Obtain the trained information matching model; The attribute information and the target smooth interaction factor are input into the trained information matching model to obtain the matching degree between the resource to be recommended and the target user.

13. The method according to claim 12, characterized in that The generation of the information matching model includes the following steps: Acquire training samples of the information matching model to be trained; the training samples include training smooth interaction factors corresponding to training recommendation resources, attribute information of training users, and matching labels of training recommendation resources and training users; Taking the training smooth interaction factor corresponding to the training recommendation resource and the attribute information of the training user as inputs of the information matching model, and obtaining the predicted matching result of the training recommendation resource and the training user; The model parameters of the information matching model are adjusted based on the predicted matching results and the matching labels until a convergence condition is met, thereby obtaining a trained information matching model.

14. A resource recommendation device, characterized in that: The device comprises: An information acquisition module is used to acquire target interaction information and target display information of the resource to be recommended; and to acquire historical interaction information and historical display information of the historically recommended resource; A factor determination module, used to determine a target interaction factor and a target display factor based on the target interaction information, the target display information, the historical interaction information and the historical display information; An information smoothing module, used to perform sample smoothing processing according to the target interaction factor and the target display factor to obtain the target smooth interaction factor corresponding to the resource to be recommended; A matching degree determination module is used to obtain attribute information of a target user, and to perform information matching based on the attribute information and a target smooth interaction factor corresponding to the resource to be recommended, so as to obtain a matching degree between the resource to be recommended and the target user; A resource recommendation module is used to recommend the resource to be recommended according to the matching degree.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 13 are implemented.

16. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

Citation Information

Patent Citations

  • Advertisement data recommendation method and system

    CN107590689A

  • Advertisement recommending method, user equipment and server

    CN107895278A

  • Information processing method and device, computer equipment and storage medium

    CN108133013A

  • Resource recommendation method and device and electronic equipment

    CN110413877A

  • Predictive recommendation system

    US10325285B1