Information processing method, device, computer equipment and storage medium

Abstract

The present application relates to an information processing method, apparatus, computer device, and storage medium. The method comprises: obtaining relative ranking information corresponding to target feature information in various information recommendation models; obtaining weight information corresponding to each of the information recommendation models; determining comprehensive ranking information of the target feature information based on the relative ranking information corresponding to the target feature information in various information recommendation models and the weight information corresponding to each of the information recommendation models; and determining the importance level of the target feature information based on the comprehensive ranking information of the target feature information. This method can accurately identify the importance of feature information.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to an information processing method, apparatus, computer equipment, and storage medium. Background Art

[0002] In traditional internet technology, websites recommend potential business objects to users, enabling them to quickly find the ones they need from the recommended objects. For example, shopping websites might recommend products similar to those users searched or browsed for, while news websites might recommend news similar to those users searched or browsed for. Recommended business objects are determined by recommendation models, and the accuracy of these models is related to the characteristics of the sample data used by the recommendation models.

[0003] Generally speaking, businesses need to maintain feature information. However, when a business has too many recommendation models, the amount of feature information is also enormous, which places a heavy workload on the business. Businesses can prioritize the maintenance of important feature information, but traditional technologies cannot accurately identify the importance of feature information. Summary of the Invention

[0004] Based on this, it is necessary to provide an information processing method, device, computer equipment and storage medium that can accurately identify the importance of feature information in order to address the above technical problems.

[0005] An information processing method, the method comprising:

[0006] Obtaining the relative ranking information corresponding to the target feature information in each information recommendation model. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The sample feature information is used to train the information recommendation model.

[0007] Obtain the weight information corresponding to each information recommendation model;

[0008] Determine the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model.

[0009] The importance level of the target feature information is determined based on the comprehensive ranking information of the target feature information.

[0010] An information processing device, comprising:

[0011] An acquisition module is used to obtain the relative ranking information corresponding to the target feature information in each information recommendation model. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The sample feature information is used to train the information recommendation model.

[0012] The acquisition module is also used to obtain the weight information corresponding to each information recommendation model;

[0013] A determination module is used to determine comprehensive ranking information of the target feature information based on the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model;

[0014] The determination module is further used to determine the importance level of the target feature information based on the comprehensive ranking information of the target feature information.

[0015] In one embodiment, the acquisition module is further used to: detect the number of times the information recommendation model is online within a preset time period; when the number of times the information recommendation model is online within the preset time period is once, obtain the relative ranking information corresponding to the target feature information in the information recommendation model this time, and use the relative ranking information corresponding to the target feature information in the information recommendation model this time as the relative ranking information corresponding to the target feature information in the information recommendation model.

[0016] In one embodiment, the acquisition module is also used to: when the information recommendation model is online at least twice within a preset time period, obtain the relative ranking information of the target feature information corresponding to at least two online accesses of the information recommendation model; and determine the relative ranking information corresponding to the target feature information in the information recommendation model based on the relative ranking information of the target feature information corresponding to at least two online accesses of the information recommendation model.

[0017] In one embodiment, the acquisition module is also used to: input the relative ranking information corresponding to the target feature information in at least two online information recommendation models into a preset mean function for calculation, to obtain the relative ranking information corresponding to the target feature information in the information recommendation model.

[0018] In one embodiment, the acquisition module is further used to: obtain ranking information corresponding to the target feature information in each information recommendation model, where the ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model; obtain the number of sample feature information of each information recommendation model; and determine the relative ranking information corresponding to the target feature information in each information recommendation model based on the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model.

[0019] In one embodiment, the acquisition module is further used to: obtain the contribution degree corresponding to the target feature information in each information recommendation model; and determine the ranking information corresponding to the target feature information in each information recommendation model based on the contribution degree corresponding to the target feature information in each information recommendation model.

[0020] In one embodiment, the acquisition module is further used to: obtain the number of recommendations and the number of conversions corresponding to each information recommendation model; and determine the weight information corresponding to each information recommendation model based on the number of recommendations and the number of conversions corresponding to each information recommendation model.

[0021] In one embodiment, the acquisition module is further used to: obtain the product between the number of recommendations and the number of conversions corresponding to each information recommendation model, and use the product between the number of recommendations and the number of conversions corresponding to each information recommendation model as the weight information corresponding to each information recommendation model.

[0022] In one embodiment, the determination module is further used to: obtain the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model; input the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model into a predetermined mean function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0023] In one embodiment, the determination module is further used to: obtain the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model; input the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model into a preset extreme value function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0024] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are performed:

[0025] Obtaining the relative ranking information corresponding to the target feature information in each information recommendation model. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The sample feature information is used to train the information recommendation model.

[0026] Obtain the weight information corresponding to each information recommendation model;

[0027] Determine the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model.

[0028] The importance level of the target feature information is determined based on the comprehensive ranking information of the target feature information.

[0029] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0030] Obtaining the relative ranking information corresponding to the target feature information in each information recommendation model. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The sample feature information is used to train the information recommendation model.

[0031] Obtain the weight information corresponding to each information recommendation model;

[0032] Determine the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model.

[0033] The importance level of the target feature information is determined based on the comprehensive ranking information of the target feature information.

[0034] The above-mentioned information processing method, device, computer equipment and storage medium obtain the relative ranking information corresponding to the target feature information in each information recommendation model, obtain the weight information corresponding to each information recommendation model, determine the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to the target feature information in each information recommendation model, and determine the importance level of the target feature information based on the comprehensive ranking information of the target feature information. In this way, the relative ranking information corresponding to the feature information in an information recommendation model is first obtained, and then the comprehensive ranking information corresponding to the feature information in each information recommendation model is determined based on the relative ranking information, thereby determining the importance level of the feature information based on the comprehensive ranking information, thereby achieving accurate identification of the importance of the feature information. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A diagram of an application environment of an information processing method in one embodiment;

[0036] Figure 2 1 is a flow chart of an information processing method in one embodiment;

[0037] Figure 3 A schematic diagram of a detailed process for obtaining relative ranking information in one embodiment;

[0038] Figure 4 A schematic diagram of a detailed process for obtaining relative ranking information in another embodiment;

[0039] Figure 5 A schematic diagram of a detailed process for obtaining relative ranking information in yet another embodiment;

[0040] Figure 6 A schematic diagram of a detailed process for obtaining ranking information in one embodiment;

[0041] Figure 7 A schematic diagram of a detailed process for obtaining weight information in one embodiment;

[0042] Figure 8 A schematic diagram of a detailed process for obtaining comprehensive ranking information in one embodiment;

[0043] Figure 9 A schematic diagram of a detailed process for obtaining comprehensive ranking information in another embodiment;

[0044] Figure 10 is a structural block diagram of an information processing device in one embodiment;

[0045] Figure 11 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0047] In internet technology, enterprises need to maintain feature information. However, when an enterprise has too many recommendation models, the amount of feature information is also enormous, which places a heavy workload on the enterprise. Enterprises can prioritize the maintenance of important feature information. For example, feature information can be restructured and optimized based on its importance level. Based on a preset level, feature information with an importance level higher than or equal to the preset level is considered important, while feature information with an importance level lower than the preset level is considered unimportant. Important feature information can be brought online while unimportant information can be taken offline, thereby optimizing the overall structure of feature information and reducing the workload of feature information maintenance.

[0048] The information processing method provided in the present application first obtains the relative ranking information corresponding to the feature information in an information recommendation model, and then determines the comprehensive ranking information corresponding to the feature information in each information recommendation model based on the relative ranking information, thereby determining the importance level of the feature information based on the comprehensive ranking information, thereby accurately identifying the importance of the feature information.

[0049] The information processing method provided in this application can be applied to Figure 1 In the application environment shown. The computer device 102 communicates with the server 104 via a network. Specifically, the computer device 102 obtains relative ranking information corresponding to the target feature information in each information recommendation model through the server 104. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model, and the sample feature information is used to train the information recommendation model. The computer device 102 obtains weight information corresponding to each information recommendation model through the server 104. The computer device 102 determines comprehensive ranking information of the target feature information based on the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model. The computer device 102 determines the importance level of the target feature information based on the comprehensive ranking information of the target feature information. The computer device 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The server 104 can be implemented as an independent server or a server cluster consisting of multiple servers.

[0050] In one embodiment, Figure 2 As shown, an information processing method is provided. The information processing method can be executed by a computer device. This embodiment uses the method executed by a computer device as an example. It is understandable that the method can also be executed by a server, or by a system including a computer device and a server, and implemented through the interaction between the computer device and the server. In this embodiment, the method includes the following steps:

[0051] Step 202: Obtain the relative ranking information corresponding to the target feature information in each information recommendation model. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The sample feature information is used to train the information recommendation model.

[0052] The information recommendation model is used to determine the information to be recommended based on the acquired information.

[0053] In one embodiment, the category to which the user identifier belongs is obtained, and the recommended information is determined based on the category to which the user identifier belongs. The user identifier uniquely identifies the user and can be a user ID number, passport number, driver's license number, etc. The category to which the user identifier belongs can be determined based on the characteristic information corresponding to the user identifier. The characteristic information describes the behavior and status of the user corresponding to the user identifier in certain areas.

[0054] In one embodiment, historical browsing information corresponding to the user identifier is obtained, information similar to the historical browsing information is determined, and the similar information is recommended to the user identifier. Alternatively, information currently being browsed by the user identifier is obtained, information similar to the currently being browsed information is determined, and the similar information is recommended to the user identifier.

[0055] Information recommendation models can be used in a variety of business scenarios. For example, they can be used to recommend news published by news websites, financial products published by financial websites, products listed on shopping websites, videos published by video websites, music listed on music websites, books uploaded by book websites, accounts on social networking sites, feeds, and so on.

[0056] The target feature information is one of the sample feature information, and the target feature information is the feature information of the sample to be tested. The sample feature information is selected from the sample data. The sample data is used to train the information recommendation model. Optionally, the sample data is labeled using label information, and the sample feature information is selected from the sample data. The sample feature information is input into the information recommendation model, and the information recommendation model processes the sample feature information to obtain a prediction result. The parameters of the information recommendation model are adjusted based on the difference between the prediction result and the label information corresponding to the sample feature information.

[0057] The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The relative ranking information of the target feature information can characterize the importance of the target feature information in the sample feature information of the information recommendation model. In one embodiment, obtaining the relative ranking information corresponding to the target feature information in each information recommendation model includes: obtaining the ranking information corresponding to the target feature information in each information recommendation model, and determining the relative ranking information corresponding to the target feature information in each information recommendation model based on the ranking information corresponding to the target feature information in each information recommendation model, wherein the ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model.

[0058] In one embodiment, the relative ranking information corresponding to the target feature information in each information recommendation model may be determined based on the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model.

[0059] In one embodiment, the computer device obtains relative ranking information corresponding to the target feature information in each information recommendation model.

[0060] Step 204: Obtain weight information corresponding to each information recommendation model.

[0061] Among them, the weight information corresponding to each information recommendation model can represent the importance of each information recommendation model.

[0062] In one embodiment, different information recommendation models can be used for different business scenarios, and the weight information corresponding to the information recommendation model can be determined according to the importance of the business scenario. For example, the business scenario corresponding to information recommendation model A is more important than the business scenario corresponding to information recommendation model B, so the weight information corresponding to information recommendation model A is higher than the weight information corresponding to information recommendation model B. Alternatively, the business scenario corresponding to information recommendation model A is equally important as the business scenario corresponding to information recommendation model B, so the weight information corresponding to information recommendation model A is equal to the weight information corresponding to information recommendation model B, and the weight information corresponding to information recommendation model A and information recommendation model B can be preset values, such as 1. In one embodiment, the mapping relationship between the information recommendation model and the weight information can be pre-set, and the weight information corresponding to the information recommendation model can be determined based on the information recommendation model and the mapping relationship.

[0063] In one embodiment, the computer device obtains weight information corresponding to each information recommendation model.

[0064] Step 206: Determine the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model.

[0065] The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model. The comprehensive ranking information of the target feature information can represent the importance of the target feature information in the sample feature information of each information recommendation model.

[0066] In one embodiment, the computer device determines the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model.

[0067] Step 208: Determine the importance level of the target feature information based on the comprehensive ranking information of the target feature information.

[0068] Among them, the importance level is used to represent the importance of the target feature information in the sample feature information of each information recommendation model. The higher the importance level, the more important the target feature information is.

[0069] In one embodiment, a mapping relationship between comprehensive ranking information and importance levels is preset, and the importance level of the target feature information is determined based on the comprehensive ranking information of the target feature information and the mapping relationship.

[0070] In one embodiment, feature information can be restructured and optimized based on its importance level. For example, based on a preset level, when the importance level of feature information is higher than or equal to the preset level, the feature information is considered important, and when the importance level of feature information is lower than the preset level, the feature information is considered unimportant. Important feature information can be brought online, while unimportant feature information can be taken offline, thereby optimizing the overall structure of the feature information and reducing the workload of maintaining the feature information.

[0071] In one embodiment, the computer device determines the importance level of the target feature information based on the comprehensive ranking information of the target feature information.

[0072] In the above-mentioned information processing method, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, and the weight information corresponding to each information recommendation model is obtained. According to the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, the comprehensive ranking information of the target feature information is determined, and the importance level of the target feature information is determined based on the comprehensive ranking information of the target feature information. In this way, the relative ranking information corresponding to the feature information in an information recommendation model is first obtained, and then the comprehensive ranking information corresponding to the feature information in each information recommendation model is determined based on the relative ranking information, so that the importance level of the feature information is determined based on the comprehensive ranking information, thereby achieving accurate identification of the importance of the feature information.

[0073] In one embodiment, Figure 3 As shown, the method for obtaining the relative ranking information corresponding to the target feature information in the information recommendation model includes:

[0074] Step 302: Detect the number of times the information recommendation model is online within a preset time period.

[0075] Among them, the preset duration can be set according to actual application, such as 1 month, 6 months, 12 months, etc.

[0076] Specifically, difficult sample data for the information recommendation model can be collected, labeled using labeling information, and then used to train the information recommendation model to optimize the model's parameters. Difficult sample data can be sample data that the information recommendation model incorrectly identified during testing or online use. The information recommendation model, optimized based on the difficult sample data, needs to be re-launched.

[0077] In one embodiment, the information recommendation model may be optimized and trained periodically, for example, every 10 days. Alternatively, when a certain amount of difficult sample data is accumulated, the information recommendation model may be optimized and trained.

[0078] Step 304: When the information recommendation model is online once within the preset time period, the relative ranking information of the target feature information corresponding to the current online information recommendation model is obtained, and the relative ranking information of the target feature information corresponding to the current online information recommendation model is used as the relative ranking information corresponding to the target feature information in the information recommendation model.

[0079] Specifically, when the information recommendation model is online once within a preset time period, the relative ranking information corresponding to the target feature information in the information recommendation model can be directly used as the relative ranking information corresponding to the target feature information in the information recommendation model.

[0080] In this embodiment, the number of times the information recommendation model is online within a preset time period is detected. When the number of times the information recommendation model is online within the preset time period is once, the relative ranking information corresponding to the target feature information in the information recommendation model this time is obtained, and the relative ranking information corresponding to the target feature information in the information recommendation model this time is used as the relative ranking information corresponding to the target feature information in the information recommendation model. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0081] In one embodiment, Figure 4 As shown, the method for obtaining the relative ranking information corresponding to the target feature information in the information recommendation model includes:

[0082] Step 402: Detect the number of times the information recommendation model is online within a preset time period.

[0083] Step 404: When the information recommendation model is online at least twice within a preset time period, relative ranking information corresponding to the target feature information when the information recommendation model is online at least twice is obtained.

[0084] Specifically, when the information recommendation model is online at least twice within a preset time period, since the sample data used in each training is different, the relative ranking information corresponding to the target feature information at least twice when the information recommendation model is online is also different. Therefore, it is necessary to obtain the relative ranking information corresponding to the target feature information at least twice when the information recommendation model is online.

[0085] Step 406 , determining the relative ranking information corresponding to the target feature information in the information recommendation model based on the relative ranking information corresponding to at least two online appearances of the target feature information in the information recommendation model.

[0086] In one embodiment, the maximum value or the minimum value in the relative ranking information corresponding to at least two online visits of the target feature information in the information recommendation model is used as the relative ranking information corresponding to the target feature information in the information recommendation model.

[0087] In this embodiment, the number of times the information recommendation model goes online within a preset time period is detected. When the number of times the information recommendation model goes online within the preset time period is at least twice, the relative ranking information corresponding to the target feature information at least twice being online in the information recommendation model is obtained. Based on the relative ranking information corresponding to the target feature information at least twice being online in the information recommendation model, the relative ranking information corresponding to the target feature information in the information recommendation model is determined. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0088] In one embodiment, the relative ranking information corresponding to the target feature information in the information recommendation model is determined based on the relative ranking information corresponding to at least two times the target feature information is online in the information recommendation model, including: inputting the relative ranking information corresponding to at least two times the target feature information is online in the information recommendation model into a preset mean function for calculation, and obtaining the relative ranking information corresponding to the target feature information in the information recommendation model.

[0089] Among them, the preset mean function can be used to calculate the average value.

[0090] In one embodiment, the relative ranking information corresponding to at least two online accesses of the target feature information in the information recommendation model is input into a preset mean function for calculation to obtain the average value of the relative ranking information corresponding to at least two online accesses of the target feature information in the information recommendation model, and the average value of the relative ranking information corresponding to at least two online accesses of the target feature information in the information recommendation model is used as the relative ranking information corresponding to the target feature information in the information recommendation model. For example, taking information recommendation model A as an example, the number of online accesses of information recommendation model A within the preset time length is X times, the relative ranking information corresponding to the first online access of the target feature information in information recommendation model A is Y1, the relative ranking information corresponding to the second online access of the target feature information in information recommendation model A is Y2, ..., the relative ranking information corresponding to the Xth online access of the target feature information in information recommendation model A is Y X , then the relative ranking information corresponding to the target feature information in the information recommendation model A is: (Y1+Y2+……+Y X ) / X.

[0091] In one embodiment, corresponding weights can be set for the target feature information that appears online at least twice in the information recommendation model, and the product between the relative ranking information corresponding to the target feature information that appears online at least twice in the information recommendation model and the corresponding weight is calculated. The product between the relative ranking information corresponding to the target feature information that appears online at least twice in the information recommendation model and the corresponding weight is input into a preset mean function for calculation to obtain a weighted average of the relative ranking information corresponding to the target feature information that appears online at least twice in the information recommendation model, and the weighted average of the relative ranking information corresponding to the target feature information that appears online at least twice in the information recommendation model is used as the relative ranking information corresponding to the target feature information in the information recommendation model. For example, taking information recommendation model A as an example, the number of times information recommendation model A appears online within a preset time period is X times, the relative ranking information corresponding to the first time the target feature information appears online in information recommendation model A is Y1, and the weight corresponding to the first time is Z1, the relative ranking information corresponding to the second time the target feature information appears online in information recommendation model A is Y2, and the weight corresponding to the second time is Z2, ..., the relative ranking information corresponding to the Xth time the target feature information appears online in information recommendation model A is Y X , the weight corresponding to the Xth time of going online is Z X , then the relative ranking information corresponding to the target feature information in the information recommendation model A is: (Z 1* Y1+Z 2* Y2+……+Z X* Y X ) / X.

[0092] In this embodiment, the relative ranking information corresponding to the target feature information in at least two online information recommendation models is input into a preset mean function for calculation to obtain the relative ranking information corresponding to the target feature information in the information recommendation model. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0093] In one embodiment, Figure 5 As shown, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, including:

[0094] Step 502: Obtain ranking information corresponding to the target feature information in each information recommendation model. The ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model.

[0095] Among them, the ranking information is used to describe the ranking of the target feature information in the sample feature information used in this training of the information recommendation model.

[0096] In one embodiment, the ranking information corresponding to the target feature information in each information recommendation model can be determined using a random forest algorithm. A random forest is a classifier that contains multiple decision trees, and its output category is determined by the mode of the categories output by the individual trees. Specifically, a forest is randomly constructed with multiple decision trees. Each decision tree in the random forest is independent of the others. When a sample is input, each decision tree determines the category to which the sample belongs. The category that is selected most frequently is the category to which the sample belongs.

[0097] In one embodiment, the ranking information corresponding to the target feature information in each information recommendation model can be determined by the difference method of out-of-bag sample prediction errors. Specifically, in the random forest, when constructing each decision tree, a different Bootstrap Sample (randomly and with replacement) is used for the training set. Therefore, for each decision tree (assuming that for the Pth decision tree), a part of the training instances does not participate in the generation of the Pth decision tree, and this part of the training instances is the out-of-bag sample of the Pth decision tree. For each decision tree in the random forest, the out-of-bag sample is used to calculate its out-of-bag sample error rate, which is recorded as Erroob1; noise interference is randomly added to the feature information of the out-of-bag sample, and its out-of-bag sample error rate is calculated again, which is recorded as Erroob2; assuming that there are Q decision trees in the random forest, the importance of the feature information is: ∑(Erroob2-Erroob1) / Q. If the out-of-bag sample error rate increases significantly after adding noise to a certain feature information, it means that the feature information has a great influence on the classification result of the sample data, that is, the feature information is more important, and the higher the ranking of the feature information is.

[0098] Step 504: Obtain the number of sample feature information of each information recommendation model. The number of sample feature information of the information recommendation model refers to the number of sample feature information used in the current training of the information recommendation model.

[0099] Step 506: Determine the relative ranking information corresponding to the target feature information in each information recommendation model based on the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model.

[0100] In one embodiment, the quotient of the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model is obtained, and the quotient of the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model is used as the relative ranking information corresponding to the target feature information in each information recommendation model. For example, taking information recommendation model A as an example, the ranking information corresponding to the target feature information in information recommendation model A is M1, and the number of sample feature information of information recommendation model A is N1, then the relative ranking information corresponding to the target feature information in information recommendation model A is M1 / N1.

[0101] In this embodiment, the ranking information corresponding to the target feature information in each information recommendation model is obtained, the number of sample feature information of each information recommendation model is obtained, and the relative ranking information corresponding to the target feature information in each information recommendation model is determined based on the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0102] In one embodiment, Figure 6 As shown, the ranking information corresponding to the target feature information in each information recommendation model is obtained, including:

[0103] Step 602: Obtain the contribution of the target feature information in each information recommendation model.

[0104] Among them, contribution is used to characterize the importance of target feature information in the training process of the information recommendation model.

[0105] In one embodiment, the contribution of the target feature information in each information recommendation model can be calculated by the mean_decrease_impurity method, and the average value of the impurity reduction of each decision tree can be calculated as the contribution of the feature information. In the mean impurity reduction method, for classification problems, Gini impurity or information gain can be used for calculation; for regression problems, variance or least squares fitting can be used for calculation. Specifically, in the splitting of each decision tree, the accumulation of the degree of improvement of a feature information on the segmentation criterion is used to measure the importance of the feature information. The average value is taken over all decision trees to obtain the overall importance of the feature information, which is the contribution of the feature information.

[0106] For a single decision tree T k ,k=1,2,…,k, calculate the importance of target feature information U in a single decision tree:

[0107]

[0108] Among them, J is the decision tree T k The number of split nodes in ,λ t 2 Represents the improvement of the segmentation criterion on the node. The I(V(t)=U) function indicates that if the segmentation feature of the node is U, the value is 1, otherwise the value is 0.

[0109] In one embodiment, the contribution of target feature information in each information recommendation model can be calculated using a coefficient method. Specifically, all feature information can be normalized, and a regression model or an attention model can be established for the normalized feature information. The absolute value of the regression coefficient can be obtained, and the proportion of the absolute value of the regression coefficient can be calculated to obtain the contribution of the feature information.

[0110] Step 604: Determine the ranking information corresponding to the target feature information in each information recommendation model based on the contribution of the target feature information in each information recommendation model.

[0111] In one embodiment, the target feature information may be sorted from high to low according to its corresponding contribution in each information recommendation model to obtain ranking information corresponding to the target feature information in each information recommendation model.

[0112] In this embodiment, the contribution degree corresponding to the target feature information in each information recommendation model is obtained, and the ranking information corresponding to the target feature information in each information recommendation model is determined based on the contribution degree corresponding to the target feature information in each information recommendation model. In this way, the ranking information corresponding to the target feature information in each information recommendation model is accurately obtained.

[0113] In one embodiment, Figure 7 As shown, obtain the weight information corresponding to each information recommendation model, including:

[0114] Step 702: Obtain the number of recommendations and conversions corresponding to each information recommendation model.

[0115] The number of recommendations refers to the number of user visits to the business scenario corresponding to the information recommendation model within the predetermined time period. The number of conversions can be the number of user clicks on the information recommended by the information recommendation model within the predetermined time period, or the number of user conversions on the information recommended by the information recommendation model within the predetermined time period. The predetermined time period can be 1 day. For example, taking the application of the information recommendation model to a news application as an example, the average daily number of user visits to the news application is 1 million, and for one of the news items, the average daily number of user clicks is 500. Then the number of recommendations corresponding to the information recommendation model is 1 million, and the number of conversions corresponding to the information recommendation model is 500.

[0116] Step 704: Determine the weight information corresponding to each information recommendation model based on the number of recommendations and the number of conversions corresponding to each information recommendation model.

[0117] In one embodiment, the sum of the number of recommendations and the number of conversions corresponding to each information recommendation model may be obtained, and the sum of the number of recommendations and the number of conversions corresponding to each information recommendation model may be used as the weight information corresponding to each information recommendation model.

[0118] In one embodiment, weights can be set for the number of recommendations and the number of conversions, and a first product between the number of recommendations corresponding to each information recommendation model and the corresponding weight, as well as a second product between the number of conversions corresponding to each information recommendation model and the corresponding weight, are obtained. The sum of the first product and the second product is obtained, and the sum of the first product and the second product is used as the weight information corresponding to each information recommendation model.

[0119] In one embodiment, the amount of revenue can be determined based on the number of conversions, and the weight information corresponding to each information recommendation model can be determined based on the number of recommendations and the amount of revenue corresponding to each information recommendation model. Optionally, the method for determining the amount of revenue based on the number of conversions can be: weighting the number of user clicks and the number of user conversions, obtaining a third product between the number of user clicks and the corresponding weight, obtaining a fourth product between the number of user conversions and the corresponding weight, obtaining the sum of the third product and the fourth product, and using the sum of the third product and the fourth product as the amount of revenue; or using the third product or the fourth product as the amount of revenue.

[0120] In one embodiment, the number of times each information recommendation model goes online within a preset time period is detected. When the number of times the information recommendation model goes online within the preset time period is at least twice, the weight information corresponding to the target feature information at least twice when the information recommendation model goes online is obtained. Based on the weight information corresponding to the target feature information at least twice when the information recommendation model goes online, the weight information corresponding to the information recommendation model is determined.

[0121] In one embodiment, the weight information corresponding to at least two instances of the target feature information appearing online in the information recommendation model is input into a preset mean function for calculation, obtaining an average value of the weight information corresponding to at least two instances of the target feature information appearing online in the information recommendation model, and using the average value as the weight information corresponding to the information recommendation model. Alternatively, the weight information corresponding to at least two instances of the target feature information appearing online in the information recommendation model is input into a predetermined extreme value function for calculation, obtaining a maximum or minimum value of the weight information corresponding to at least two instances of the target feature information appearing online in the information recommendation model, and using the maximum or minimum value as the weight information corresponding to the information recommendation model.

[0122] In this embodiment, the number of recommendations and the number of conversions corresponding to each information recommendation model are obtained, and the weight information corresponding to each information recommendation model is determined based on the number of recommendations and the number of conversions corresponding to each information recommendation model, so that the importance of the information recommendation model can be accurately identified.

[0123] In one embodiment, the weight information corresponding to each information recommendation model is determined based on the number of recommendations and the number of conversions corresponding to each information recommendation model, including: obtaining the product of the number of recommendations and the number of conversions corresponding to each information recommendation model, and using the product of the number of recommendations and the number of conversions corresponding to each information recommendation model as the weight information corresponding to each information recommendation model.

[0124] In this embodiment, the product of the number of recommendations and the number of conversions corresponding to each information recommendation model is obtained, and the product of the number of recommendations and the number of conversions corresponding to each information recommendation model is used as the weight information corresponding to each information recommendation model, so that the importance of the information recommendation model can be accurately identified.

[0125] In one embodiment, Figure 8 As shown, based on the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, the comprehensive ranking information of the target feature information is determined, including:

[0126] Step 802: Obtain the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model.

[0127] In step 804, the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model is input into a predetermined mean function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0128] A predetermined mean function may be used to calculate the average value.

[0129] In one embodiment, the relative ranking information corresponding to the target feature information in each information recommendation model and the product of the weight information corresponding to each information recommendation model are obtained, and the relative ranking information corresponding to the target feature information in each information recommendation model and the product of the weight information corresponding to each information recommendation model are input into a predetermined mean function for calculation to obtain the average value of the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, and the average value is used as the comprehensive relative ranking information of the target feature information.

[0130] In this embodiment, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, and the product of the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model is input into a predetermined mean function for calculation to obtain the comprehensive relative ranking information of the target feature information. In this way, the comprehensive relative ranking information of the target feature information can be accurately identified.

[0131] In one embodiment, Figure 9 As shown, based on the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, the comprehensive ranking information of the target feature information is determined, including:

[0132] Step 902: Obtain the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model.

[0133] In step 904, the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model is input into a preset extreme value function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0134] Among them, the preset extreme value function can be used to calculate the maximum or minimum value.

[0135] In one embodiment, the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model is input into a preset extreme value function for calculation to obtain the minimum value of the product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, and the minimum value is used as the comprehensive relative ranking information of the target feature information.

[0136] In this embodiment, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, and the product of the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model is input into a preset extreme value function for calculation to obtain the comprehensive relative ranking information of the target feature information. In this way, the comprehensive relative ranking information of the target feature information can be accurately identified.

[0137] The present application also provides an application scenario, which applies the above-mentioned information processing method. Specifically, the application of the information processing method in this application scenario is as follows: according to the importance level of the feature information, the feature information is reconstructed and optimized. Taking the preset level as the boundary, when the importance level of the feature information is higher than or equal to the preset level, the feature information is important feature information, and when the importance level of the feature information is lower than the preset level, the feature information is unimportant feature information. Important feature information can be put online and unimportant feature information can be taken offline, thereby optimizing the overall structure of the feature information and reducing the workload of maintaining the feature information. The method includes:

[0138] Obtain ranking information corresponding to the target feature information in each information recommendation model. The ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model. Obtain the number of sample feature information of each information recommendation model. According to the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model, determine the relative ranking information corresponding to the target feature information in each information recommendation model.

[0139] Next, the number of recommendations and conversions corresponding to each information recommendation model is obtained, and the weight information corresponding to each information recommendation model is determined based on the number of recommendations and conversions corresponding to each information recommendation model. The product of the number of recommendations and conversions corresponding to each information recommendation model can be obtained, and the product of the number of recommendations and conversions corresponding to each information recommendation model is used as the weight information corresponding to each information recommendation model.

[0140] Furthermore, based on the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, comprehensive ranking information of the target feature information is determined. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model. The product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model can be obtained. The product of the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model is input into a predetermined mean function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0141] Next, the importance level of the target feature information is determined based on the comprehensive ranking information of the target feature information.

[0142] In the above information processing method, the relative ranking information corresponding to the feature information in an information recommendation model is first obtained, and then the comprehensive ranking information corresponding to the feature information in each information recommendation model is determined based on the relative ranking information, and then the importance level of the feature information is determined based on the comprehensive ranking information, so as to accurately identify the importance of the feature information.

[0143] It should be understood that although Figure 2-8 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. In addition, Figure 2-8 At 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 order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0144] In one embodiment, Figure 10 As shown, an information processing device is provided. The device can be a software module or a hardware module, or a combination of the two to form a part of a computer device. The device specifically includes: an acquisition module 1002 and a determination module 1004, wherein:

[0145] Acquisition module 1002, for acquiring relative ranking information corresponding to target feature information in various information recommendation models. The relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model. The sample feature information is used to train the information recommendation model.

[0146] The acquisition module 1002 is further used to obtain weight information corresponding to each information recommendation model;

[0147] Determination module 1004 is used to determine comprehensive ranking information of the target feature information based on the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model. The comprehensive ranking information is used to describe the ranking of the target feature information in the sample feature information of each information recommendation model;

[0148] The determination module 1004 is further configured to determine the importance level of the target feature information according to the comprehensive ranking information of the target feature information.

[0149] In the above-mentioned information processing device, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, and the weight information corresponding to each information recommendation model is obtained. According to the relative ranking information corresponding to the target feature information in each information recommendation model and the weight information corresponding to each information recommendation model, the comprehensive ranking information of the target feature information is determined, and according to the comprehensive ranking information of the target feature information, the importance level of the target feature information is determined. In this way, the relative ranking information corresponding to the feature information in an information recommendation model is first obtained, and then the comprehensive ranking information corresponding to the feature information in each information recommendation model is determined according to the relative ranking information, so that the importance level of the feature information is determined according to the comprehensive ranking information, thereby achieving accurate identification of the importance of the feature information.

[0150] In one embodiment, the acquisition module 1002 is further used to: detect the number of times the information recommendation model is online within a preset time period; when the number of times the information recommendation model is online within the preset time period is once, obtain the relative ranking information corresponding to the target feature information in the information recommendation model this time, and use the relative ranking information corresponding to the target feature information in the information recommendation model this time as the relative ranking information corresponding to the target feature information in the information recommendation model.

[0151] In this embodiment, the number of times the information recommendation model is online within a preset time period is detected. When the number of times the information recommendation model is online within the preset time period is once, the relative ranking information corresponding to the target feature information in the information recommendation model this time is obtained, and the relative ranking information corresponding to the target feature information in the information recommendation model this time is used as the relative ranking information corresponding to the target feature information in the information recommendation model. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0152] In one embodiment, the acquisition module 1002 is further used to: when the information recommendation model is online at least twice within a preset time period, obtain the relative ranking information of the target feature information corresponding to at least two online accesses of the information recommendation model; and determine the relative ranking information corresponding to the target feature information in the information recommendation model based on the relative ranking information of the target feature information corresponding to at least two online accesses of the information recommendation model.

[0153] In this embodiment, when the information recommendation model is online at least twice within a preset time period, the relative ranking information corresponding to the target feature information at least twice being online in the information recommendation model is obtained. Based on the relative ranking information corresponding to the target feature information at least twice being online in the information recommendation model, the relative ranking information corresponding to the target feature information in the information recommendation model is determined. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0154] In one embodiment, the acquisition module 1002 is further used to: input the relative ranking information corresponding to the target feature information in at least two online information recommendation models into a preset mean function for calculation, to obtain the relative ranking information corresponding to the target feature information in the information recommendation model.

[0155] In this embodiment, the relative ranking information corresponding to the target feature information in at least two online information recommendation models is input into a preset mean function for calculation to obtain the relative ranking information corresponding to the target feature information in the information recommendation model. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0156] In one embodiment, the acquisition module 1002 is further used to: obtain ranking information corresponding to the target feature information in each information recommendation model, where the ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model; obtain the number of sample feature information of each information recommendation model; and determine the relative ranking information corresponding to the target feature information in each information recommendation model based on the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model.

[0157] In this embodiment, the ranking information corresponding to the target feature information in each information recommendation model is obtained, the number of sample feature information of each information recommendation model is obtained, and the relative ranking information corresponding to the target feature information in each information recommendation model is determined based on the ranking information corresponding to the target feature information in each information recommendation model and the number of sample feature information of each information recommendation model. In this way, the relative ranking information corresponding to the target feature information in the information recommendation model is accurately obtained.

[0158] In one embodiment, the acquisition module 1002 is further used to: obtain the contribution degree of the target feature information in each information recommendation model; and determine the ranking information of the target feature information in each information recommendation model based on the contribution degree of the target feature information in each information recommendation model.

[0159] In this embodiment, the contribution degree corresponding to the target feature information in each information recommendation model is obtained, and the ranking information corresponding to the target feature information in each information recommendation model is determined based on the contribution degree corresponding to the target feature information in each information recommendation model. In this way, the ranking information corresponding to the target feature information in each information recommendation model is accurately obtained.

[0160] In one embodiment, the acquisition module 1002 is further used to: obtain the number of recommendations and the number of conversions corresponding to each information recommendation model; and determine the weight information corresponding to each information recommendation model based on the number of recommendations and the number of conversions corresponding to each information recommendation model.

[0161] In this embodiment, the number of recommendations and the number of conversions corresponding to each information recommendation model are obtained, and the weight information corresponding to each information recommendation model is determined based on the number of recommendations and the number of conversions corresponding to each information recommendation model, so that the importance of the information recommendation model can be accurately identified.

[0162] In one embodiment, the acquisition module 1002 is further used to: obtain the product between the number of recommendations and the number of conversions corresponding to each information recommendation model, and use the product between the number of recommendations and the number of conversions corresponding to each information recommendation model as the weight information corresponding to each information recommendation model.

[0163] In this embodiment, the product of the number of recommendations and the number of conversions corresponding to each information recommendation model is obtained, and the product of the number of recommendations and the number of conversions corresponding to each information recommendation model is used as the weight information corresponding to each information recommendation model, so that the importance of the information recommendation model can be accurately identified.

[0164] In one embodiment, the determination module 1004 is further used to: obtain the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model; input the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model into a predetermined mean function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0165] In this embodiment, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, and the product of the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model is input into a predetermined mean function for calculation to obtain the comprehensive relative ranking information of the target feature information. In this way, the comprehensive relative ranking information of the target feature information can be accurately identified.

[0166] In one embodiment, the determination module 1004 is further used to: obtain the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model; input the relative ranking information corresponding to the target feature information in each information recommendation model, and the product of the weight information corresponding to each information recommendation model into a preset extreme value function for calculation to obtain the comprehensive relative ranking information of the target feature information.

[0167] In this embodiment, the relative ranking information corresponding to the target feature information in each information recommendation model is obtained, and the product of the relative ranking information corresponding to each information recommendation model and the weight information corresponding to each information recommendation model is input into a preset extreme value function for calculation to obtain the comprehensive relative ranking information of the target feature information. In this way, the comprehensive relative ranking information of the target feature information can be accurately identified.

[0168] For the specific definition of the information processing device, please refer to the definition of the information processing method above and will not be repeated here. Each module in the above-mentioned information processing device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned 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 each of the above modules.

[0169] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 11As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. 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, an information processing 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 provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0170] Those skilled in the art will understand that Figure 11 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 shown in the figure, or combine certain components, or have a different component arrangement.

[0171] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0172] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0173] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented 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 may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may 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).

[0174] The technical features of the above embodiments can 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.

[0175] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. An information processing method, characterized in that: The method comprises: Detect the number of times each information recommendation model is online within a preset time period; When the information recommendation model is online at least twice within a preset time period: corresponding weights are set for the target feature information when it is online at least twice on the information recommendation model, and the product of the relative ranking information corresponding to the target feature information when it is online at least twice on the information recommendation model and the corresponding weight is input into a preset mean function to calculate, and a weighted average of the relative ranking information corresponding to the target feature information when it is online at least twice on the information recommendation model is obtained as the relative ranking information corresponding to the target feature information in the information recommendation model, the relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model, and the sample feature information is used to train the information recommendation model; Setting weights for the number of user clicks and the number of user conversions, performing weighted calculations on the number of user clicks and the number of user conversions to obtain the amount of revenue; Determine the weight information corresponding to each information recommendation model according to the number of recommendations corresponding to each information recommendation model and the amount of revenue; Obtaining the product of relative ranking information corresponding to the target feature information in the information recommendation model and weight information corresponding to the information recommendation model; Inputting the product into a predetermined function for calculation to obtain comprehensive relative ranking information of the target feature information; the comprehensive relative ranking information is used to describe the ranking of the target feature information in the sample feature information of each of the information recommendation models; The importance level of the target feature information is determined according to the comprehensive relative ranking information of the target feature information.

2. The method according to claim 1, characterized in that The method further comprises: When the information recommendation model is online once within the preset time period, the relative ranking information corresponding to the target feature information in the information recommendation model this time is obtained, and the relative ranking information corresponding to the target feature information in the information recommendation model this time is used as the relative ranking information corresponding to the target feature information in the information recommendation model; the weight information corresponding to each of the information recommendation models is obtained; and the comprehensive ranking information of the target feature information is determined according to the relative ranking information corresponding to the target feature information in each of the information recommendation models and the weight information corresponding to each of the information recommendation models.

3. The method according to claim 1, characterized in that The method further comprises: The weight information corresponding to the target feature information when the information recommendation model is online for at least two times is input into a preset mean function or extreme value function for calculation, and the average value, maximum value or minimum value of the weight information corresponding to the target feature information when the information recommendation model is online for at least two times is obtained, and the average value, maximum value or minimum value is used as the weight information corresponding to the information recommendation model.

4. The method according to claim 1, wherein The method further comprises: Obtaining ranking information corresponding to each online launch of the target feature information in each of the information recommendation models, wherein the ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model; Obtaining the quantity of sample feature information of each of the information recommendation models; According to the ranking information corresponding to each of the information recommendation models each time the target feature information goes online, and the number of sample feature information of each of the information recommendation models, the relative ranking information corresponding to each of the information recommendation models each time the target feature information goes online is determined.

5. The method according to claim 4, characterized in that The obtaining of ranking information corresponding to each online launch of each information recommendation model for the target feature information includes: Obtaining the contribution of the target feature information to each of the information recommendation models each time they are launched; According to the contribution of the target feature information in each of the information recommendation models each time it goes online, the ranking information of the target feature information in each of the information recommendation models each time it goes online is determined.

6. The method according to claim 1, characterized in that The method further comprises: Obtain the number of recommendations and conversions corresponding to each of the information recommendation models; The weight information corresponding to each of the information recommendation models is determined according to the recommendation quantity and the conversion quantity corresponding to each of the information recommendation models.

7. The method according to claim 6, characterized in that The determining, based on the recommendation quantity and the conversion quantity corresponding to each information recommendation model, weight information corresponding to each information recommendation model includes: The product of the number of recommendations and the number of conversions corresponding to each of the information recommendation models is obtained, and the product of the number of recommendations and the number of conversions corresponding to each of the information recommendation models is used as the weight information corresponding to each of the information recommendation models.

8. The method according to claim 1, characterized in that The predetermined function is a mean function or an extreme value function.

9. An information processing device, characterized in that The device comprises: An acquisition module is used to detect the number of times each information recommendation model goes online within a preset time period; when the number of times the information recommendation model goes online within the preset time period is at least twice: corresponding weights are set for the target feature information when it goes online at least twice in the information recommendation model, and the product of the relative ranking information corresponding to the target feature information when it goes online at least twice in the information recommendation model and the corresponding weight is input into a preset mean function calculation to obtain the weighted average of the relative ranking information corresponding to the target feature information when it goes online at least twice in the information recommendation model, as the relative ranking information corresponding to the target feature information in the information recommendation model, the relative ranking information is used to describe the relative ranking of the target feature information in the sample feature information of the information recommendation model, and the sample feature information is used to train the information recommendation model; weights are set for the number of user clicks and the number of user conversions, and the number of user clicks and the number of user conversions are weightedly calculated to obtain the amount of revenue; the weight information corresponding to each information recommendation model is determined according to the number of recommendations corresponding to each information recommendation model and the amount of revenue; a calculation module, configured to obtain the product of the relative ranking information corresponding to the target feature information in the information recommendation model and the weight information corresponding to the information recommendation model; input the product into a predetermined function for calculation to obtain comprehensive relative ranking information of the target feature information; the comprehensive relative ranking information is used to describe the ranking of the target feature information in the sample feature information of each of the information recommendation models; The determination module is used to determine the importance level of the target feature information according to the comprehensive relative ranking information of the target feature information.

10. The information processing device according to claim 9, wherein The acquisition module is also used for: when the information recommendation model has been online once within the preset time period, obtaining the relative ranking information corresponding to the target feature information in the information recommendation model this time, and using the relative ranking information corresponding to the target feature information in the information recommendation model this time as the relative ranking information corresponding to the target feature information in the information recommendation model; obtaining the weight information corresponding to each of the information recommendation models; and determining the comprehensive ranking information of the target feature information based on the relative ranking information corresponding to the target feature information in each of the information recommendation models and the weight information corresponding to each of the information recommendation models.

11. The information processing device according to claim 9, wherein The acquisition module is also used to: input the weight information corresponding to the target feature information when the information recommendation model is online at least twice into a preset mean function or extreme value function for calculation, obtain the average value or maximum value or minimum value of the weight information corresponding to the target feature information when the information recommendation model is online at least twice, and use the average value or the maximum value or the minimum value as the weight information corresponding to the information recommendation model.

12. The information processing device according to claim 9, wherein The acquisition module is also used to: obtain the ranking information corresponding to each of the target feature information each time the information recommendation model goes online, the ranking information is used to describe the ranking of the target feature information in the sample feature information of the information recommendation model; obtain the number of sample feature information of each of the information recommendation models; determine the relative ranking information corresponding to each of the information recommendation models each time the target feature information goes online based on the ranking information corresponding to each of the information recommendation models each time, and the number of sample feature information of each of the information recommendation models.

13. The information processing device according to claim 12, wherein: The acquisition module is also used to: obtain the contribution of the target feature information each time the information recommendation model goes online; and determine the ranking information of the target feature information each time the information recommendation model goes online based on the contribution of the target feature information each time the information recommendation model goes online.

14. The information processing device according to claim 9, wherein The acquisition module is further used to: obtain the number of recommendations and the number of conversions corresponding to each of the information recommendation models; and determine the weight information corresponding to each of the information recommendation models based on the number of recommendations and the number of conversions corresponding to each of the information recommendation models.

15. The information processing device according to claim 14, wherein: The acquisition module is also used to: obtain the product between the recommendation quantity and the conversion quantity corresponding to each of the information recommendation models, and use the product between the recommendation quantity and the conversion quantity corresponding to each of the information recommendation models as the weight information corresponding to each of the information recommendation models.

16. The information processing device according to claim 9, wherein The predetermined function is a mean function or an extreme value function.

17. 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 method according to any one of claims 1 to 8 is implemented.

18. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

19. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images