Resource display method and device, storage medium and computer device

Abstract

The application discloses a resource display method and device, a storage medium and a computer device. The method can use user characteristics, resource characteristics and scene characteristics as reference bases for selecting target resources, so that target users can obtain resources that meet their own needs, improve the use effect of the resources, help improve the total transaction amount, provide appropriate preferential policies for the users, improve the shopping experience of the users, and further enhance the loyalty of the users. Moreover, a tree structure of target feature classification is constructed based on feature engineering, and target features with significant influence are selected from multiple feature dimensions through split gain, so as to assist in determining the discrimination rules on the business and improving the interpretability of the model output results, and the accurate prediction of the purchase behavior of the target users to different resources in different business scenes is effectively realized.

Description

Technical Field

[0001] This application relates to the field of Internet technology, and in particular to a method, apparatus, storage medium, and computer device for displaying resources. Background Technology

[0002] With the rapid and continuous development of e-commerce on the internet and mobile internet, e-commerce platforms frequently issue various coupons to attract and incentivize users. However, for customers who don't frequently use coupons, these promotions are difficult for them to discover and utilize, making random coupon distribution unlikely to achieve the desired promotional effect. Especially for paid coupon services, where users need to purchase before using, traditional coupon distribution strategies cannot accurately recommend coupons that meet user needs, thus failing to maximize the return on investment for coupons, given the business objective of maximizing coupon sales. Summary of the Invention

[0003] In view of this, this application provides a method, apparatus, storage medium, and computer device for displaying resources, which selects suitable resources based on user characteristics, resource characteristics, and scenario characteristics, thereby accurately recommending resources to users to improve the platform's total transaction volume and unique visitor metrics.

[0004] According to one aspect of this application, a method for displaying a resource is provided, comprising:

[0005] In response to the order placed by the target user, obtain the resource purchase information of the first user, who is a user who has made a resource purchase during a preset time period;

[0006] The user characteristics and resource purchase information of the first user are subjected to feature filtering processing to determine the sample user characteristics, common resource characteristics and common scenario characteristics of the first user and the second user. The second user is a user who does not have resource purchase operations within a preset time period.

[0007] A decision tree model is constructed by performing feature cross-processing on the characteristics of sample users, common resource features, and common scenario features.

[0008] Based on the split gain of the split nodes in the decision tree model, the target features in the sample user features, common resource features, and common scene features, as well as the correlation between the target features, are determined.

[0009] Machine learning operations are performed based on target features and relationships to obtain a purchase probability model;

[0010] Based on the output of the purchase probability model, display the purchase control for the target resource corresponding to the output result on the target user's order interface.

[0011] Optionally, the user features and the resource purchase information of the first user are subjected to feature screening processing, specifically including:

[0012] According to the user features of the first user and the second user, sample user features and feature weights of the sample user features are determined.

[0013] The resource features and the scene features in the resource purchase information are subjected to cleaning processing respectively, and common resource features and common scene features of a third user are determined, the third user being a first user corresponding to the sample user features.

[0014] The first difference degree of the common resource features with respect to the second user and the second difference degree of the common scene features with respect to the second user are determined.

[0015] Optionally, according to the user features of the first user and the second user, sample user features and feature weights of the sample user features are determined, specifically including:

[0016] If the quantity difference between the first user and the second user corresponding to the user features is greater than a preset difference value, the user features are determined as difference user features.

[0017] The feature weights of the difference user features are determined.

[0018] The difference user features are subjected to screening processing according to the feature weights, and sample user features are obtained.

[0019] Optionally, the feature weights of the difference user features are determined, specifically including:

[0020] According to the proportion of the first user corresponding to the difference user features in the sample users, the feature weights are calculated; and / or,

[0021] According to the resource purchase information, the correlation coefficients of the difference user features with the first user and the second user are determined respectively, and the feature weights are calculated according to the correlation coefficients.

[0022] Optionally, the sample user features include first features and / or second features; the difference user features are subjected to screening processing according to the feature weights, and the sample user features are determined, specifically including:

[0023] If the feature weight of the difference user features is greater than a preset feature weight, the difference user features are determined as the first features.

[0024] If the quantity of the first features is multiple, the multiple first features are subjected to combination processing, and the second features are obtained.

[0025] Optionally, the resource features and the scene features in the resource purchase information are subjected to cleaning processing respectively, specifically including:

[0026] If the proportion of the third users corresponding to the resource feature in the first users corresponding to the resource feature is greater than or equal to a first preset ratio, the resource feature is determined as a common resource feature;

[0027] If the proportion of the third users corresponding to the scene feature in the first users corresponding to the scene feature is greater than or equal to a second preset ratio, the scene feature is determined as a common scene feature.

[0028] Optionally, the first difference degree of the common resource feature with respect to the second user and the second difference degree of the common scene feature with respect to the second user are determined, specifically including:

[0029] The first difference degree of the common resource feature is determined according to the number of the third users corresponding to the second user and the common resource feature;

[0030] The second difference degree of the common scene feature is determined according to the number of the third users corresponding to the second user and the common scene feature.

[0031] Optionally, the output result includes the target resource feature, the target scene feature and the purchase probability which are associated with each other; and the purchase control of the target resource corresponding to the output result is displayed on the order interface of the target user according to the output result of the purchase probability model, specifically including:

[0032] If the preset resource contains the target resource feature, the product of the price of the preset resource and the purchase probability is calculated as the estimated conversion amount corresponding to the target resource feature;

[0033] If the estimated conversion amount corresponding to the target resource feature is greater than or equal to a first threshold value, and the purchase probability is greater than or equal to a second threshold value, the preset resource containing the target resource feature is determined as the target resource;

[0034] If the scene feature of the business scene in which the target user is located meets the target scene feature, the purchase control of the target resource containing the target resource feature associated with the target scene feature is displayed on the order interface.

[0035] According to another aspect of the present application, a resource display device is provided, including:

[0036] An acquisition module is configured to acquire resource purchase information of a first user in response to an order operation of a target user, the first user being a user who has a resource purchase operation in a preset period;

[0037] A feature determination module is configured to perform feature screening processing on user features and resource purchase information of the first user, and determine sample user features between the first user and a second user, common resource features and common scene features corresponding to the first user, the second user being a user who has no resource purchase operation in the preset period;

[0038] The constructing module is configured to perform feature cross processing on the sample user features, the common resource features, and the common scene features to construct a decision tree model.

[0039] The associating module is configured to determine target features in the sample user features, the common resource features, and the common scene features, and an association relationship between the target features according to a split gain of a split node in the decision tree model.

[0040] The training module is configured to perform a machine learning operation according to the target features and the association relationship to obtain a purchase probability model.

[0041] The displaying module is configured to display a purchase control of a target resource corresponding to an output result of the purchase probability model on an order interface of a target user according to the output result.

[0042] Optionally, the feature determining module is specifically configured to determine sample user features and feature weights of the sample user features according to user features of a first user and a second user; perform cleaning processing on resource features and scene features in resource purchase information respectively to determine common resource features and common scene features of a third user, the third user being the first user corresponding to the sample user features; and determine a first difference degree of the common resource features relative to the second user and a second difference degree of the common scene features relative to the second user.

[0043] Optionally, the feature determining module includes:

[0044] The user feature module is configured to determine a user feature as a difference user feature if a quantity difference between the first user and the second user corresponding to the user feature is greater than a preset difference value, and determine a feature weight of the difference user feature.

[0045] The screening module is configured to perform screening processing on the difference user features according to the feature weights to obtain the sample user features.

[0046] Optionally, the sample user features include first features and / or second features.

[0047] The screening module is specifically configured to determine a difference user feature as a first feature if a feature weight of the difference user feature is greater than a preset feature weight, and perform combination processing on a plurality of first features to obtain second features if a quantity of the first features is a plurality.

[0048] Optionally, the feature determining module is specifically configured to determine a resource feature as a common resource feature if a quantity proportion of a third user corresponding to the resource feature in a first user corresponding to the resource feature is greater than or equal to a first preset proportion, and determine a scene feature as a common scene feature if a quantity proportion of a third user corresponding to the scene feature in a first user corresponding to the scene feature is greater than or equal to a second preset proportion.

[0049] Optionally, the feature determination module is specifically configured to determine a first difference degree of the common resource feature according to a number of third users corresponding to the second user and the common resource feature; and determine a second difference degree of the common scene feature according to a number of third users corresponding to the second user and the common scene feature.

[0050] Optionally, the output result comprises the target resource feature, the target scene feature and the purchase probability which are associated with each other; and the resource display device further comprises:

[0051] The estimation module is configured to calculate a product of a price and the purchase probability of the preset resource as an estimated conversion amount corresponding to the target resource feature if the preset resource comprises the target resource feature; and determine the preset resource comprising the target resource feature as the target resource if the estimated conversion amount corresponding to the target resource feature is greater than or equal to a first threshold value and the purchase probability is greater than or equal to a second threshold value.

[0052] The display module is specifically configured to display a purchase control of the target resource comprising the target resource feature associated with the target scene feature on the order interface if the scene feature of the business scene in which the target user is located meets the target scene feature.

[0053] By the technical scheme, when the target user performs the order placing operation of the target commodity, it indicates that the target user may have a purchase resource to deduct the consumption demand. At this time, the preset user is classified to determine the first user who has a resource purchase operation in a preset period and the second user who does not have a resource purchase operation in the preset period, and the resource purchase information of the resource purchased by the first user is obtained. The sample user features that are different between the two types of users and the common features of resources and business scenarios (common resource features and common scenario features) between different first users are determined through feature screening processing. Then, the sample user features, the common resource features and the common scenario features corresponding to the two different types of users are subjected to feature cross processing, and a decision tree model is constructed to determine the target features in the sample user features, the common resource features and the common scenario features, and the association relationship between the target features, so as to determine the feature contribution degree of the data in each feature dimension to the resource purchase of the user. Then, the machine learning operation is performed according to the target features and the association relationship, and a purchase probability model is obtained, so that the purchase probability model can reflect the purchase preference difference of the two different types of users to the multiple resources in the multiple business scenarios, and the difference is used to display the estimated purchase control of the target resource on the order interface of the target user. Therefore, the target user can obtain the resource suitable for the consumption demand of the target user through the purchase control before submitting the order. Through the above technical scheme, on the one hand, the user features, resource features and scenario features are used as the reference basis for selecting the target resource, so that the target user can obtain the resource more suitable for the user's own demand, which not only improves the use effect of the resource, but also helps to improve the total transaction amount, provides appropriate preferential policies for the user, improves the user shopping experience, and further enhances the user loyalty. On the other hand, the tree structure of the target feature classification is constructed based on the feature engineering, and the target features with significant influence are selected from multiple feature dimensions through split gain, to assist in determining the discrimination rule in business, improve the interpretability of the model output result, and effectively realize the accurate estimation of the purchase behavior of the target user to different resources in different business scenarios.

[0054] The above description is only a summary of the technical scheme of the present application. In order to more clearly understand the technical means of the present application, the content of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0055] The drawings described herein are used to provide further understanding of the present application, and form a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0056] Figure 1A flowchart of a method for displaying resources is shown.

[0057] Figure 2 A structural block diagram of a device for displaying resources is shown. DETAILED DESCRIPTION

[0058] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0059] The embodiments of the present application will be described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar notations represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as a limitation of the present application.

[0060] Those skilled in the art can understand that the singular forms "a", "an" and "the" used herein include plural forms unless specifically stated otherwise. It should be further understood that the use of the term "include" in the specification of the present application means that the features, integers, steps, operations, elements and / or components described therein are present, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we say an element is "connected" or "joined" to another element, it can be directly connected or joined to the other element, or there can be an intermediate element. In addition, "connected" or "joined" used herein can include wireless connection or wireless connection. The phrase "and / or" used herein includes all or any single unit and all combinations of the associated listed items.

[0061] Now, exemplary embodiments according to the present application will be described in more detail with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in various different forms, and should not be interpreted as being limited only to the embodiments set forth herein. It should be understood that these embodiments are provided so that the disclosure of the present application is complete and complete, and the concepts of these exemplary embodiments are fully conveyed to those skilled in the art.

[0062] In the present embodiment, a method for displaying resources is provided, as shown in the figure, the method comprises: Figure 1

[0063] Step 101, in response to the ordering operation of the target user, acquiring the resource purchase information of the first user;

[0064] In the present embodiment, a method for displaying resources is provided, as shown in the figure, the method comprises:

[0065] ​Specifically, the resource purchase information includes a purchase time of the resource, a purchased resource feature, a scenario feature of a business scenario in which the purchase operation is performed, and the like. The resource can be a coupon, a token, a red envelope, and the like. The resource feature includes, but is not limited to, at least one of the following: a resource type, a resource price, a resource threshold, a resource quota, a resource discount, a resource validity period, and a maximum use period. The scenario feature includes, but is not limited to, at least one of the following: a station area feature, a consumption heat, and a POI (Point of Interest) feature around the station.

[0066] In an actual application scenario, before the resource purchase information of the first user is acquired, the resource display method further includes: if the preset user has a resource purchase operation within a preset time period, determining the preset user as the first user; and if the preset user does not have a resource purchase operation within the preset time period, determining the preset user as the second user.

[0067] It can be understood that the preset user is a user with reference value that is preset, for example, a user who has logged in an e-commerce platform within a year or a user who frequently consumes, and the embodiments of the present application will not be listed one by one.

[0068] In this embodiment, the resource purchase record of the preset user is called, and whether the preset user has a resource purchase operation within the preset time period is determined based on this. If the preset user performs a resource purchase operation, it indicates that the demand of the preset user for the resource is relatively high, and the preset user can be divided into the first user with the purchase operation at this time. Conversely, if the preset user does not perform a resource purchase operation within the preset time, it indicates that the demand of the preset user for the resource is relatively low, and the preset user can be divided into the second user without the purchase operation. In order to determine the resource purchase difference of different types of users by using the classified first user and second user, and then achieve the purpose of accurately recommending the resource.

[0069] In step 102, the user features and the resource purchase information of the first user are subjected to feature screening processing, to determine sample user features between the first user and the second user, a common resource feature corresponding to the first user, and a common scenario feature.

[0070] The second user is a user who does not have a resource purchase operation within a preset time period.

[0071] In this embodiment, the sample user features that are different between the two types of users are determined through the feature screening process, and the commonalities of the resources and the business scenarios in which the different first users are located (common resource features and common scenario features) are determined. In order to select the target resources that can be recommended by using the user features, the resource features, and the scenario features, the target users can obtain resources that are more in line with their own needs, which can not only improve the use effect of the resources, but also help to increase the total transaction amount, provide appropriate preferential policies for the users, improve the shopping experience of the users, and further enhance the loyalty of the users.

[0072] In an actual application scenario, step 102, that is, the feature screening process is performed on the user features and the resource purchase information of the first user, specifically includes the following steps.

[0073] Step 102-1, determining the sample user features and the feature weights of the sample user features according to the user features of the first user and the second user.

[0074] The user features are data related to user consumption, for example, the user features include but are not limited to at least one of the following: consumption ability level, price sensitivity, loyalty, resource demand intensity, scenario preference feature, resource preference feature, and user behavior data (such as click, browse, and stay time) for paid resources.

[0075] In this embodiment, by comparing the user features of the first user and the second user, the sample user features that are different between the two types of users are determined, and the feature weights of the sample user features with respect to the resource purchase operation. Thus, through a large amount of statistical and analysis work, an index that is obviously associated with the resources and the purchase operation is found, so that the index can effectively represent the user's tendency to purchase resources for shopping.

[0076] For example, the user rarely shops online or is not used to using coupons, and even if the user is recommended a coupon, the user will not purchase the coupon, and the user is likely to be annoyed. Therefore, after the user features of the first user with the purchase operation and the second user without the purchase operation are subdivided and reorganized, it is found that there are 100 first users with high consumption ability level and low price sensitivity, and only 5 second users, and the number of different types of users is greatly different, so that the sample user features (consumption ability level and price sensitivity) are strongly correlated. When the two indexes meet certain conditions, the probability of the coupon being purchased is very low. For example, the consumption ability level of a user in the shoe category is level 1 (not often buying shoes above 500 yuan), and if the threshold of the shoe coupon is 1000 yuan, the probability of the user purchasing the coupon for consumption deduction is low. For another example, user A is not sensitive to the price of some categories of goods (rarely uses coupons), and according to this information, it can be known that even if user A is recommended a coupon, the probability of the coupon being purchased is low.

[0077] Further, as a refinement and extension of the foregoing embodiment, in order to fully describe the specific implementation process of the embodiment, step 102-1, i.e., determining the sample user features and the feature weights of the sample user features according to the user features of the first user and the second user, specifically includes:

[0078] Step 102-1-a, if the quantity difference between the first user and the second user corresponding to the user feature is greater than the preset difference, determining the user feature as a difference user feature;

[0079] Step 102-1-b, determining the feature weight of the difference user feature;

[0080] Specifically, the feature weight of the difference user feature can be determined in the following manner:

[0081] Manner one, calculating the feature weight according to the proportion of the first user corresponding to the difference user feature in the preset users.

[0082] In this embodiment, the difference user feature with a large difference between the first user and the second user is determined according to the quantity difference between the first user and the second user corresponding to the same user feature. For any difference user feature, the proportion of the first user corresponding to any difference user feature in the preset users is substituted into the preset calculation algorithm to calculate the feature weight. In order to screen out the user features that have a greater impact on the purchase users and the non-purchase users from the plurality of difference user features, these important user features can be used as the basic data for subsequent analysis. The calculation algorithm can be used to represent the corresponding relationship between the proportion and the feature weight, which can be reasonably set as needed, and the present embodiment does not make specific limitations.

[0083] Manner two, determining the correlation coefficients of the difference user features with the first user and the second user respectively according to the resource purchase information, and calculating the feature weight according to the correlation coefficients.

[0084] In this embodiment, if the screened difference user features are multiple, there may be a large error in calculating the feature weight by the proportion. Therefore, the correlation coefficients of the difference user features with the first user and the second user are introduced to calculate the feature weight, so as to ensure that the feature weight can more accurately reflect the contribution of each difference user feature.

[0085] Specifically, the correlation coefficient includes at least one of the following: Pearson correlation coefficient (Person), Spearman correlation coefficient (Spearman), and Kendall correlation coefficient (Kendall tau). According to formula (1), the Pearson correlation coefficient can be obtained:

[0086]

[0087] wherein p represents a Pearson correlation coefficient, X and Y represent the feature data related to the difference user features corresponding to the first user and the second user respectively, con(X, Y) represents the covariance of X and Y, σ X represents the standard deviation of X, and σ Y represents the standard deviation of Y.

[0088] According to the formula (2), the Spearman correlation coefficient S can be obtained.

[0089]

[0090] wherein S represents a Spearman correlation coefficient, d i represents the difference of the ranks of the two observed variables (the first user and the second user), wherein d i = x i -y i , and n represents the descending order of the original data according to the average in the overall data.

[0091] According to the formula (3), the Kendall correlation coefficient γ can be obtained.

[0092]

[0093] wherein γ represents a Kendall correlation coefficient, the number of concordant pairs represents the concordant pairs of the statistical objects according to the same order, the number of discordant pairs represents the discordant pairs of the statistical objects according to the different order, and n represents the number of the statistical objects.

[0094] Step 102-1-c, screening the difference user features according to the feature weight to obtain the sample user features.

[0095] In this embodiment, the difference user features which have a larger difference between the first user and the second user for the purchase user and the non-purchase user are screened from the plurality of user features by comparing the quantity difference between the first user and the second user of the same user feature. And the feature weight of the difference user feature relative to the resource purchase operation is determined. After determining the feature weight, the difference user features are screened according to the size of the feature weight. The difference user features with the feature weight greater than the preset feature weight are taken as the sample user features which are obviously associated with the purchase operation, so as to represent the user's tendency of purchasing the resource for shopping through the sample user features.

[0096] Further, the sample user features include the first features and / or the second features, and step 102-1-c specifically includes: if the feature weight of the difference user feature is greater than the preset feature weight, determining the difference user feature as the first feature; if the number of the first features is a plurality, performing combination processing on the plurality of first features to obtain the second feature.

[0097] In this embodiment, after determining the difference user features, the difference user features with feature weights greater than the preset feature weight are selected as the first features. If there are a plurality of first features, combination processing is performed on the plurality of first features to form new second features. Thus, the difference user features with greater difference and higher feature weight are further refined, so as to classify the users according to the important features in the user features, to calculate the commonness and difference degree of the purchased and non-purchased resource users in the resource feature or scene feature preference in the same type of user group, and then to allocate resources to each user.

[0098] For example, the 3 first features are male, preference for shoe purchase, and low price sensitivity, respectively, and after combination processing, 4 new second features are generated: male with low price sensitivity, male with preference for shoe purchase, preference for shoe purchase and low price sensitivity, and male with preference for shoe purchase and low price sensitivity. The above 3 first features and 4 second features are used as sample user features required for feature engineering.

[0099] Step 102-2, respectively performing cleaning processing on the resource features and scene features in the resource purchase information to determine the common resource features and common scene features of the third user;

[0100] The third user is the first user corresponding to the sample user features.

[0101] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to completely describe the specific implementation process of the embodiment, step 102-2, i.e., respectively performing cleaning processing on the resource features and scene features in the resource purchase information, specifically includes:

[0102] Step 102-2-a, if the number of the third users corresponding to the resource features in the first users corresponding to the resource features accounts for greater than or equal to a first preset ratio, determining the resource features as the common resource features;

[0103] Step 102-2-b, if the number of the third users corresponding to the scene features in the first users corresponding to the scene features accounts for greater than or equal to a second preset ratio, determining the scene features as the common scene features;

[0104] The first preset ratio and the second preset ratio can be reasonably set according to the feature screening accuracy.

[0105] In this embodiment, the third user is classified according to the sample user feature with a high feature weight in the user feature, and the tendency of most first users to resource-related data (common resource features and common scenario features) is determined by the proportion of the number of third users with the same sample user feature among all first users.

[0106] Taking resource features as an example, among the 100 people who redeemed coupons with high price sensitivity, 50 people purchased 10-yuan no-threshold coupons, 40 people purchased 40-yuan no-threshold coupons, and among the other 10 people, 3 purchased 88-yuan no-threshold coupons, 2 purchased category-specific coupons, and 5 purchased 1000-yuan discount coupons. According to the first preset ratio, 10-yuan no-threshold coupons and 40-yuan no-threshold coupons can be considered as common resource features. This buying or not buying is common and different, so after classifying the third users with the same sample user feature by category, we look at which coupons they have commonly bought and which coupons they have not bought, and then respectively count the features of the purchased coupons and the features of the non-purchased coupons, find the rules, and use them as the basis for subsequent model training.

[0107] Step 102-3, determining a first difference degree of the common resource features with respect to the second user and a second difference degree of the common scenario features with respect to the second user.

[0108] In this embodiment, by analyzing the resource purchase information of all first users, the resource features and scenario features are analyzed to find the common points (common resource features) and business scenario common points (common scenario features) among the first users who actually purchased resources for different sample user features. By comparing with the second user, the difference degree of the common resource features with respect to the second user and the second difference degree of the common scenario features with respect to the second user are determined, and then the contribution of different features to the user's purchase of resources is analyzed through commonality and difference, so as to subsequently allocate more suitable resources for each target user according to their own needs. This not only helps to improve the utilization rate of resources, but also saves shopping funds for users and improves user stickiness.

[0109] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully describe the specific implementation process of this embodiment, step 102-3 specifically includes:

[0110] Step 102-3-a, determining a first difference degree of the common resource features according to the number of third users corresponding to the second user and the common resource features;

[0111] Step 102-3-b, determining a second difference degree of the common scenario features according to the number of third users corresponding to the second user and the common scenario features.

[0112] In this embodiment, the third user is classified according to the sample user features with high feature weights in the user features, and the difference degrees of the users who have purchased and have not purchased the resources to different resource features in different business scenarios are calculated in the user group of each classification feature by the number of users. So as to reflect the purchase difference of the users with the same user features to various resources in various business scenarios through the difference degrees.

[0113] In step 103, the sample user features, common resource features and common scenario features are processed by feature intersection, and a decision tree model is constructed.

[0114] The decision tree model is a classical algorithm model in the field of machine learning, the decision tree is a tree structure representing classification of samples based on features, and the classification process of the decision tree can be summarized as a process of selecting the optimal partition feature recursively according to the criterion of feature selection from the given data set, and dividing the data set according to the selected optimal partition feature, so that each data subset has the best classification.

[0115] In this embodiment, the tree structure of target feature classification is constructed based on feature engineering, and the target features with significant influence are selected from multiple feature dimensions by split gain, to assist in determining the discrimination rule in business, improve the interpretability of the model output result, and effectively realize the accurate prediction of the purchase behavior of the target user to different resources in different business scenarios.

[0116] In actual application scenarios, for the decision tree model, if all features are fully crossed, the model information will be too much, which will load the model and cause information bias. Therefore, the feature intersection method can be determined by the first difference degree, the second difference degree and the feature weight, and the features that can be cross-fused are selected by the feature intersection method, and the feature splitting is performed to obtain multiple splitting results, and finally the decision tree model containing a large number of new multi-dimensional cross features is generated by using the splitting results. Thus, the requirement of feature intersection processing on the system is reduced, and the accuracy of tree model construction is improved.

[0117] It can be understood that multiple range intervals can be set for the first difference degree, the second difference degree and the feature weight respectively, the range intervals correspond to the feature intersection methods one by one, and the range intervals can be reasonably set according to the platform business requirements.

[0118] For example, if the second difference degree and the feature weight are located in the corresponding first range interval, it indicates that the common scene feature corresponding to the second difference degree has a larger difference degree relative to the second user, and the sample user feature corresponding to the feature weight also has a larger contribution degree to the user triggering the purchase operation. Therefore, it is determined that the feature intersection mode is full intersection, that is, the common scene feature and the sample user feature are fused to form a new feature containing double features. For example, for the consumption scene, whether the consumption station is a popular station or a non-popular station, the intersection of the price sensitivity of the user and the station heat will form multiple scenes, including how many resources the user with high price sensitivity purchases in the high-heat station and how many resources the user with high price sensitivity purchases in the low-heat station. Combining the user feature with the resource feature and / or the consumption scene feature forms a new feature. The more intersections, the finer the granularity, and the more details the subsequent model learns.

[0119] For another example, if the feature weights of the sample user feature A and the sample user feature B are located in the corresponding second range interval, and the contribution degree of the sample user feature corresponding to the feature weight to the user triggering the purchase operation is general, it is determined that the feature intersection mode is half intersection. At this time, the sample user feature A and the sample user feature B are displayed to enable the management personnel to select the sample user feature A and the sample user feature B, and subsequent feature intersection processing is performed according to the sample user feature selected by the management personnel, and the sample user feature not selected is stopped from splitting. Therefore, according to which type of feature is more and which type of feature is less in the analyzed features, the less abundant feature is half intersected to form a new feature. For example, the new feature generated by the intersection of the male feature in the user feature and other features is sufficient. At this time, the worker can only perform intersection and fusion of the female information as a user feature and other features to form a new feature.

[0120] For another example, if the first difference degree, the second difference degree, and the feature weight are located in the corresponding third range interval, it indicates that the common resource feature corresponding to the first difference degree has a smaller difference degree relative to the second user, and the common scene feature corresponding to the second difference degree has a smaller difference degree relative to the second user, and the contribution degree of the sample user feature corresponding to the feature weight to the user triggering the purchase operation is also lower. Therefore, it is determined that the feature intersection mode is cancel intersection, that is, the splitting of the common resource feature and the sample user feature is stopped, and a new feature will not be generated. Finally, the new features generated by full intersection and half intersection are summarized together to form a tree structure of features.

[0121] It should be noted that the generation of the decision tree model is a process of continuously dividing the data set formed by a plurality of features according to the criteria meeting the feature selection. Therefore, for the generation of the decision tree model, the training data set formed by the plurality of different dimensions of the common resource features, the common scene features and the sample user features needs to be obtained. Subsequently, each node and branch of the decision tree model is generated by continuously dividing the training data set, that is, the node splitting is continuously performed until the refined features are determined according to the data set contained in the sub-nodes obtained by the splitting, so as to finally obtain the decision tree model having the data classification function.

[0122] In step 104, the target features in the sample user features, the common resource features and the common scene features, and the association relationship between the target features are determined according to the splitting gain of the splitting node in the decision tree model.

[0123] Specifically, the decision tree model can be an Xgboost model or a GBDT model. For the GBDT model, the information gain, the Gini index and the like can be used as the splitting gain, which will not be described in detail herein.

[0124] In this embodiment, since the splitting gain is consistent with the accuracy of the model prediction, the greater the splitting gain, the higher the accuracy of the decision tree model prediction, that is, the smaller the error between the prediction value of the decision tree model and the corresponding label value. Therefore, the contribution degree of the aggregated cross features can be calculated according to the splitting gain of the splitting node (leaf) in the decision tree model, and the tree model is pruned based on this to screen out a plurality of target features with high contribution degree, and the association relationship between the target features is determined based on the pruned tree model. Thus, the granularity of the found user group is smaller, and the target features found are more accurate, which effectively improves the generalization and accuracy of the obtained decision tree model, so that subsequent personalized resource display can be performed based on the decision tree model.

[0125] It can be understood that the splitting gain can be calculated based on the label value of the feature by using the greedy algorithm (xact Greedy Algorithm). The greedy algorithm is an algorithm for finding the optimal splitting by using the exhaustive method. For example, for the user feature f1, the feature values of the user feature f1 corresponding to each common resource feature are a: 10, b: 15, c: 13, d: 12 and e: 16. First, the common resource features are sorted in ascending order of the feature values, that is, adcbe, so that six kinds of splitting results (cross features) corresponding to f1 can be obtained based on the sorting: (0, adcbe), (a, dcbe), (ad, cbe), (adc, be), (adcb, e) and (adcbe, 0), and the label value of the splitting result is calculated to calculate the splitting gain.

[0126] It is worth mentioning that the same cross feature in the decision tree model can only calculate the split gain once. However, in practical applications, due to the large number of samples and features, the probability of the same cross feature is small.

[0127] At step 105, a machine learning operation is performed according to the target feature and the association relationship to obtain a purchase probability model.

[0128] Specifically, machine learning (ML) is a multi-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, etc. Machine learning is a field of study that focuses on how computers simulate or implement human learning behavior to acquire new knowledge or skills, reorganize existing knowledge structure to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental approach to making computers intelligent. Its applications are widespread in various fields of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, etc.

[0129] At step 106, according to the output result of the purchase probability model, a purchase control of the target resource corresponding to the output result is displayed on the order interface of the target user.

[0130] The purchase control is an entrance for the user to purchase the target resource. The user can trigger the purchase control by clicking, long pressing, etc. to jump to the purchase interface of the target resource. The resource features of the target resource can be displayed on the purchase control.

[0131] It is worth mentioning that the target user represents a user who can allocate resources. Before inputting the user features of the target user into the resource display model, all users registered on the platform can be filtered through a preset condition to determine the target user who can allocate resources. For example, users with purchase records are taken as target users, or users in a login state are taken as target users. Through user filtering, the amount of data processed by the model is reduced, the system running pressure is reduced, and the display of resources to users who do not need resources is avoided, which helps to more reasonably allocate resources.

[0132] The resource display method proposed in the embodiment, when the target user performs the order placing operation of the target commodity, indicates that the target user may have the demand to purchase resources for consumption deduction. At this time, the preset user is classified to determine the first user who has the resource purchase operation in the preset time period and the second user who does not have the resource purchase operation in the preset time period, and the resource purchase information of the resources that the first user has ever purchased is obtained. The sample user features that are different between the two types of users and the common points (common resource features and common scene features) of resources and business scenes between different first users are determined through feature screening processing. Then, the sample user features, the common resource features and the common scene features corresponding to the two different types of users are subjected to feature cross processing, and a decision tree model is constructed to determine the target features in the sample user features, the common resource features and the common scene features, and the association relationship between the target features, so as to determine the feature contribution degree of the data in each feature dimension to the purchase of resources by the user. Then, the machine learning operation is performed according to the target features and the association relationship therebetween, and a purchase probability model is obtained, so that the purchase probability model can reflect the purchase preference difference of the two different types of users for various resources in various business scenes, and the difference is used to display the estimated purchase control of the target resource on the order interface of the target user. In this way, the target user can obtain resources suitable for his own consumption demand through the purchase control before submitting the order. Through the above technical solution, on the one hand, the user features, resource features and scene features are taken as the reference basis for selecting target resources, so that the target user can obtain resources that are more suitable for his own needs, which not only improves the use effect of resources, but also helps to increase the total transaction amount, provides appropriate preferential policies for users, improves the user shopping experience, and thus enhances user loyalty. On the other hand, the tree structure of target feature classification is constructed based on feature engineering, and the target features with significant influence are selected from multiple feature dimensions through split gain, to assist in determining the discrimination rules in business and improve the interpretability of the model output results, effectively realizing the accurate estimation of the purchase behavior of the target user for different resources in different business scenes.

[0133] In actual application scenarios, the output result includes the target resource feature, the target scene feature and the purchase probability; step 106, that is, displaying the purchase control of the target resource corresponding to the output result on the order interface of the target user according to the output result of the purchase probability model, specifically including the following steps:

[0134] Step 106-1, if the preset resource contains the target resource feature, calculating the product of the price of the preset resource and the purchase probability as the estimated conversion amount corresponding to the target resource feature;

[0135] In this embodiment, the user features of the target user are input into the purchase probability model to obtain target resource features suitable for the target user, target scene features, and a probability of the target user purchasing a resource under the target resource features and the target scene features. The platform side determines the purchase price of each preset resource according to business needs. After predicting the purchase probability of the target user for the possible purchase resource, the purchase probability is multiplied by the purchase price of the resource to calculate the estimated revenue that the sale of the preset resource to the target user can bring to the platform, that is, the cross merchandise volume (GMV) that the resource can convert.

[0136] In step 106-2, if the estimated conversion amount corresponding to the target resource feature is greater than or equal to a first threshold value, and the purchase probability is greater than or equal to a second threshold value, the preset resource containing the target resource feature is determined as the target resource.

[0137] In step 106-3, if the scene feature of the business scene where the target user is located matches the target scene feature, a purchase control of the target resource containing the target resource feature associated with the target scene feature is displayed on the order interface.

[0138] The first threshold value and the second threshold value can be specifically set by staff according to platform business needs.

[0139] In this embodiment, the purchase probability and the estimated revenue are taken as the basic data for subsequent judgment of whether the resource matches each user, fully considering the benefit demand of the merchant for the resource. When the estimated conversion amount corresponding to the target resource feature is greater than or equal to the first threshold value, and the purchase probability is greater than or equal to the second threshold value, it indicates that the target resource feature can not only meet the preference of the target user for the resource, but also meet the benefit demand of the merchant for the resource, so the preset resource containing the target resource feature is determined as the target resource. If the scene feature of the business scene where the target user is located matches the target scene feature, that is, the business scene where the target user is located meets the scene display condition of the resource, the target resource feature associated with the target scene feature is extracted, and a purchase control of the target resource containing the target resource feature is displayed on the order interface. Therefore, the target user can obtain the resource suitable for his own consumption demand through the purchase control before submitting the order, so as to save the shopping fund of the user. Thus, on the basis of ensuring the demand of the merchant, resources more suitable for the target user are pushed based on different business scenes and target users, realizing the optimal matching of different types of resources in different scenes for different target users. Not only is the optimal solution calculation efficiency maximized, but also the conversion rate of the resource and the user loyalty are improved, and the use experience of the user and the merchant is taken into account.

[0140] It should be noted that the target resource features obtained by the model can be one or more, and in the case of multiple target resource features, the preset resource sent to the target user needs to have multiple target resource features.

[0141] It can be understood that, in order not to affect the display of order information, the information display area of the purchase control is limited. If multiple target resources need to be displayed, it is difficult to display the features of all target resources on the purchase control. Therefore, the estimated conversion amount and the purchase probability are weighted to obtain the score of each target resource, and the multiple target resources are sorted in descending order of the score, and the features of the target resources located before the preset ranking in the ranking order are displayed on the purchase control. When the user triggers the purchase control, the purchase interface is displayed, and the features of all target resources are displayed in the purchase interface. The preset ranking can be reasonably set according to the amount of information that can be displayed by the purchase control.

[0142] For example, S1, in the single purchase scenario, when the paid coupon is displayed, the users who actually purchase and do not purchase the coupon and related data are selected.

[0143] S2, the difference features of the users who actually purchase and do not purchase the paid coupon are mined, and the feature weights of the difference features are calculated based on the real data of the historical purchase coupons of the users; the user features include user consumption ability, price sensitivity, loyalty, demand intensity, behavior characteristics, consumption scene preference, coupon preference, historical coupon use characteristics, user click, browse, and stay duration of the paid coupon, etc.

[0144] S3, according to the important features (parts with higher feature weights) in the user features, the users are classified, the classification logic includes single feature direct classification and multi-feature combination classification, and the commonness and difference degree of the coupon feature preferences of the users who purchase and do not purchase the coupon in the same type of user group are further calculated. The coupon features include type, selling price, market value, threshold, amount, discount, and validity period, etc.

[0145] S4, according to the important features (parts with higher feature weights) in the user features, the users are classified, and the commonness and difference degree of the consumption scene feature preferences of the users who purchase and do not purchase the coupon in the same type of user group are further calculated. The scene features include station area features, consumption heat, surrounding POI features, etc.

[0146] S5, a decision tree model is constructed to cross multiple features. Specifically, based on the feature weights and user difference degrees calculated in S2-S4, full crossing is performed on features with high feature weights and large difference degrees, that is, two independent features are fused into multiple new features; semi-crossing is performed on features with high feature weights and medium difference degrees, or medium feature weights and medium difference degrees, that is, two independent features are fused into one new feature; the contribution degree of the cross feature is calculated according to the split gain of the leaf node of the decision tree, and importance screening is performed to obtain target features.

[0147] S6, based on the target features in the user features, coupon features, consumption scene features, cross features and typical user historical coupon purchase data in S2-S5, a machine learning model is trained to construct a user purchase coupon probability prediction model. Through multiple rounds of iteration of model parameters, the accuracy and recall rate of predicting the user's purchase of coupons in the single purchase scene are maximized.

[0148] S7, based on the purchase probability of the user for different paid coupons in S6, the purchase price of the corresponding paid coupon is multiplied to obtain the estimated conversion GMV of the user for different paid coupons. The coupons are sorted based on the threshold 1 of the purchase probability and the threshold 2 of the user estimated conversion GMV, and the top n is taken.

[0149] S8, based on the single purchase paid coupon recommendation scene and the target user, the top n coupons of the purchase rate and the purchase conversion GMV are recommended to the corresponding target user based on the calculation results of S6 and S7.

[0150] In this embodiment, the probability of the user in the single purchase scene to purchase the paid coupon is more accurately predicted, the purchase conversion rate of the platform paid coupon is improved, and the conversion flow of the platform paid coupon is further improved.

[0151] Further, as shown in Figure 2 As a specific implementation of the resource display method described above, the embodiment of the application provides a resource display device 200, which comprises an acquisition module 201, a feature determination module 202, a construction module 203, an association module 204, a training module 205 and a display module 206.

[0152] The acquisition module 201 is configured to acquire resource purchase information of a first user in response to a placing operation of a target user, the first user being a user who has a resource purchase operation in a preset period.

[0153] The feature determination module 202 is configured to perform feature screening processing on the user features and the resource purchase information of the first user, determine sample user features between the first user and the second user, common resource features corresponding to the first user, and common scene features, and the second user is a user without resource purchase operation in a preset time period.

[0154] The construction module 203 is configured to perform feature cross processing on the sample user features, the common resource features, and the common scene features, and construct a decision tree model.

[0155] The association module 204 is configured to determine target features in the sample user features, the common resource features, and the common scene features, and an association relationship between the target features according to a split gain of a split node in the decision tree model.

[0156] The training module 205 is configured to perform a machine learning operation according to the target features and the association relationship, and obtain a purchase probability model.

[0157] The display module 206 is configured to display a purchase control of a target resource corresponding to an output result of the purchase probability model on an order interface of a target user according to the output result.

[0158] In this embodiment, when the target user performs the order operation of the target commodity, it indicates that the target user may have a purchase resource to deduct the consumption demand. At this time, the preset user classifies to determine the first user who has a resource purchase operation within a preset time period and the second user who does not have a resource purchase operation within the preset time period, and obtains the resource purchase information of the resource that the first user has ever purchased. The sample user features that are different between the two types of users and the common features of resources and business scenarios (common resource features and common scenario features) between different first users are determined through feature screening processing. Then, the sample user features, the common resource features and the common scenario features corresponding to the two different types of users are subjected to feature cross processing, and a decision tree model is constructed to determine the target features in the sample user features, the common resource features and the common scenario features, and the association relationship between the target features, so as to determine the feature contribution of the data in each feature dimension to the purchase of the resource by the user. Then, the machine learning operation is performed according to the target features and the association relationship thereof, and a purchase probability model is obtained, so that the purchase probability model can reflect the purchase preference difference of the two different types of users for multiple resources in multiple business scenarios, and the difference is used to display the estimated purchase control of the target resource on the order interface of the target user. In this way, the target user can obtain a resource suitable for his own consumption demand through the purchase control before submitting the order. Through the above technical solution, on the one hand, the user features, resource features and scenario features are taken as the reference basis for selecting the target resource, so that the target user can obtain a resource that is more suitable for his own needs, which not only improves the use effect of the resource, but also helps to increase the total transaction amount, provides appropriate preferential policies for the user, improves the user shopping experience, and thus enhances the user loyalty. On the other hand, the tree structure of the target feature classification is constructed based on feature engineering, and the target features with significant influence are selected from multiple feature dimensions through split gain, to assist in determining the discrimination rule in business and improve the explainability of the model output result, effectively realizing the accurate estimation of the purchase behavior of the target user for different resources in different business scenarios.

[0159] Further, the feature determination module 202 is specifically configured to determine the sample user features and the feature weights of the sample user features according to the user features of the first user and the second user; clean the resource features and the scenario features in the resource purchase information respectively, determine the common resource features and the common scenario features of the third user, and the third user is the first user corresponding to the sample user features; determine the first difference degree of the common resource features relative to the second user and the second difference degree of the common scenario features relative to the second user.

[0160] Further, according to the user features of the first user and the second user; the feature determination module 202 includes a user feature module (not shown in the figure) and a screening module (not shown in the figure);

[0161] The user feature module is configured to determine the user feature as a difference user feature if a quantity difference between the first user and the second user corresponding to the user feature is greater than a preset difference value, and determine a feature weight of the difference user feature.

[0162] The screening module is configured to perform screening processing on the difference user feature according to the feature weight, to obtain a sample user feature.

[0163] Further, the sample user feature includes a first feature and / or a second feature.

[0164] The screening module is specifically configured to determine the difference user feature as the first feature if the feature weight of the difference user feature is greater than a preset feature weight, and perform combination processing on a plurality of first features to obtain a second feature if the quantity of the first features is a plurality.

[0165] Further, the feature determination module 202 is specifically configured to determine the resource feature as a common resource feature if a quantity proportion of the third user corresponding to the resource feature in the first user corresponding to the resource feature is greater than or equal to a first preset proportion, and determine the scene feature as a common scene feature if a quantity proportion of the third user corresponding to the scene feature in the first user corresponding to the scene feature is greater than or equal to a second preset proportion.

[0166] Further, the feature determination module 202 is specifically configured to determine a first difference degree of the common resource feature according to the quantity of the second user and the third user corresponding to the common resource feature, and determine a second difference degree of the common scene feature according to the quantity of the second user and the third user corresponding to the common scene feature.

[0167] Further, the output result includes a target resource feature, a target scene feature and a purchase probability that are associated with each other, and the resource display device 200 further includes an estimation module (not shown in the figure).

[0168] The estimation module is configured to calculate a product of a price and the purchase probability of the preset resource as an estimated conversion amount corresponding to the target resource feature if the preset resource contains the target resource feature, and determine the preset resource containing the target resource feature as a target resource if the estimated conversion amount corresponding to the target resource feature is greater than or equal to a first threshold value and the purchase probability is greater than or equal to a second threshold value.

[0169] The display module 206 is specifically configured to display a purchase control of a target resource containing a target resource feature associated with the target scene feature on an order interface if a scene feature of a business scene in which the target user is located meets the target scene feature.

[0170] The specific limitation of the resource display device can refer to the limitation of the resource display method, which will not be repeated here. Each module in the above resource display device can be realized by software, hardware and their combination. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so that the processor calls and executes the corresponding operations of the above modules.

[0171] Based on the above method as shown in Figure 1 Accordingly, the embodiment of the present application also provides a readable storage medium, which stores a computer program, and the program is executed by a processor to realize the above resource display method as shown in Figure 1 .

[0172] Based on such understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in each embodiment scenario of the present application.

[0173] Based on the above method as shown in Figure 1 , and Figure 2 the virtual device embodiment, in order to achieve the above purpose, the embodiment of the present application also provides a computer device, which can be a personal computer, a server, a network device, etc. The computer device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to realize the above resource display method as shown in Figure 1 .

[0174] Optionally, the computer device can also include a user interface, a network interface, a camera, a radio frequency (RF) circuit, a sensor, an audio circuit, a WI-FI module, etc. The user interface can include a display screen, an input unit such as a keyboard, etc. The optional user interface can also include a USB interface, a card reader interface, etc. The network interface can optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface), etc.

[0175] Those skilled in the art can understand that the structure of the computer device provided by the embodiment does not constitute a limitation on the computer device, which can include more or fewer components, or combine certain components, or different component arrangements.

[0176] The storage medium can further include an operating system and a network communication module. The operating system is a program for managing and saving computer device hardware and software resources, supporting the running of information processing programs and other software and / or programs. The network communication module is used to realize communication between components inside the storage medium and communication with other hardware and software in the entity device.

[0177] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and a necessary general hardware platform, or by hardware to respond to the ordering operation of a target user, obtain resource purchase information of a first user, the first user being a user who has a resource purchase operation within a preset period; perform feature screening processing on the user features and resource purchase information of the first user, determine sample user features between the first user and a second user, a common resource feature corresponding to the first user and a common scene feature, the second user being a user who does not have a resource purchase operation within a preset period; perform feature cross processing on the sample user features, the common resource feature and the common scene feature to construct a decision tree model; determine target features in the sample user features, the common resource feature and the common scene feature, and the association relationship between the target features according to the split gain of the split nodes in the decision tree model; perform machine learning operation according to the target features and the association relationship to obtain a purchase probability model; and display a purchase control of a target resource corresponding to an output result of the purchase probability model on an order interface of the target user according to the output result. The embodiment of the present application takes the user features, resource features and scene features as reference bases for selecting target resources on one hand, so that the target user can obtain resources that meet his own needs, which not only improves the use effect of the resources, is conducive to improving the total transaction amount, but also provides appropriate preferential policies for the user, improves the user shopping experience, and thus enhances user loyalty. On the other hand, the tree structure of target feature classification is constructed based on feature engineering, and the target features with significant influence are selected from multiple feature dimensions based on the split gain, to assist in determining the discrimination rules in the business, improve the interpretability of the model output result, and effectively realize accurate prediction of the purchase behavior of the target user on different resources in different business scenarios.

[0178] Those skilled in the art can understand that the modules in the apparatus in the implementation scenario can be distributed in the apparatus in the implementation scenario according to the description of the implementation scenario, or can be changed and located in one or more apparatuses different from the implementation scenario. The modules of the above implementation scenario can be combined into one module, or can be further split into multiple sub-modules.

[0179] The above application number is only for description, and does not represent the advantages and disadvantages of the implementation scene. The above disclosure is only some specific implementation scenes of the application, but the application is not limited thereto, and any changes that can be thought of by those skilled in the art should fall within the protection scope of the application.

Claims

1. A method for displaying resources, characterized in that, The method includes: In response to the order placed by the target user, obtain the resource purchase information of the first user, where the first user is a user who has made a resource purchase during a preset time period; The user characteristics and resource purchase information of the first user are subjected to feature filtering processing to determine the sample user characteristics, common resource characteristics and common scenario characteristics of the first user and the second user. The second user is a user who has not made any resource purchase operations during the preset time period. The sample user features, common resource features, and common scene features are subjected to feature cross-processing to construct a decision tree model; Based on the split gain of the split nodes in the decision tree model, the target features among the sample user features, the common resource features, and the common scene features, as well as the correlation between the target features, are determined. Machine learning operations are performed based on the target features and the correlation to obtain a purchase probability model; Based on the output of the purchase probability model, the purchase control for the target resource corresponding to the output result is displayed on the order interface of the target user; The feature filtering process for the first user's user characteristics and resource purchase information specifically includes: Based on the user characteristics of the first user and the second user, determine the sample user characteristics and the feature weights of the sample user characteristics; The resource features and scenario features in the resource purchase information are cleaned and processed respectively to determine the common resource features and common scenario features of the third user, wherein the third user is the first user corresponding to the sample user features; Determine the first degree of difference of the common resource features relative to the second user and the second degree of difference of the common scenario features relative to the second user.

2. The resource display method according to claim 1, characterized in that, The step of determining the sample user features and their feature weights based on the user features of the first user and the second user specifically includes: If the difference in the number of the first user and the second user corresponding to the user feature is greater than a preset difference, the user feature is determined as the differential user feature. Determine the feature weights of the differentiated user features; The differential user features are filtered according to the feature weights to obtain the sample user features.

3. The resource display method according to claim 2, characterized in that, The sample user characteristics include a first characteristic and / or a second characteristic; the step of filtering the differential user characteristics according to the feature weights to determine the sample user characteristics specifically includes: If the feature weight of the differential user feature is greater than the preset feature weight, the differential user feature is determined as the first feature; If there are multiple first features, the multiple first features are combined to obtain the second feature.

4. The resource display method according to claim 1, characterized in that, The step of cleaning the resource features and scenario features in the resource purchase information specifically includes: If the proportion of the third user corresponding to the resource feature among the first user corresponding to the resource feature is greater than or equal to a first preset ratio, the resource feature is determined as the common resource feature; If the proportion of the third user corresponding to the scene feature among the first user corresponding to the scene feature is greater than or equal to a second preset ratio, the scene feature is determined as the common scene feature.

5. The resource display method according to claim 1, characterized in that, Determining the first degree of difference of the common resource features relative to the second user and the second degree of difference of the common scene features relative to the second user specifically includes: Based on the number of the second user and the number of the third user corresponding to the common resource feature, a first degree of difference of the common resource feature is determined; The second degree of difference of the common scene feature is determined based on the number of the second user and the number of the third user corresponding to the common scene feature.

6. The method for displaying resources according to any one of claims 1 to 5, characterized in that, The output results include interrelated target resource features, target scenario features, and purchase probabilities; the step of displaying the purchase control for the target resource corresponding to the output results on the target user's order interface based on the output results of the purchase probability model specifically includes: If the preset resource contains the target resource feature, calculate the product of the price of the preset resource and the purchase probability as the estimated conversion amount corresponding to the target resource feature; If the estimated conversion amount corresponding to the target resource feature is greater than or equal to the first threshold, and the purchase probability is greater than or equal to the second threshold, the preset resource containing the target resource feature is determined as the target resource; If the scenario characteristics of the business scenario in which the target user is located match the target scenario characteristics, a purchase control for the target resource containing the target resource characteristics associated with the target scenario characteristics will be displayed on the order interface.

7. A resource display device, characterized in that, The device includes: The acquisition module is used to respond to the order operation of the target user and acquire the resource purchase information of the first user, where the first user is a user who has a resource purchase operation within a preset time period. The feature determination module is used to perform feature filtering processing on the user features and resource purchase information of the first user, and determine the sample user features, common resource features and common scenario features of the first user and the second user. The second user is a user who has not made any resource purchase operations during the preset time period. The construction module is used to perform feature cross-processing on the sample user features, the common resource features, and the common scene features to construct a decision tree model; The association module is used to determine the target features among the sample user features, the common resource features, and the common scene features, as well as the association relationships between the target features, based on the split gain of the split nodes in the decision tree model. The training module is used to perform machine learning operations according to the target features and the correlation to obtain a purchase probability model; The display module is used to display the purchase control of the target resource corresponding to the output result on the order interface of the target user based on the output result of the purchase probability model; The feature determination module is specifically used for: Based on the user characteristics of the first user and the second user, determine the sample user characteristics and the feature weights of the sample user characteristics; The resource features and scenario features in the resource purchase information are cleaned and processed respectively to determine the common resource features and common scenario features of the third user, wherein the third user is the first user corresponding to the sample user features; Determine the first degree of difference of the common resource features relative to the second user and the second degree of difference of the common scenario features relative to the second user.

8. The resource display device according to claim 7, characterized in that, The user characteristics of the first user and the second user are used; The feature determination module includes: The user feature module is configured to determine the user feature as the differential user feature if the difference in the number of the first user and the second user corresponding to the user feature is greater than a preset difference; and... Determine the feature weights of the differentiated user features; The filtering module is used to filter the differential user features according to the feature weights to obtain the sample user features.

9. The resource display device according to claim 7, characterized in that, The sample user characteristics include a first feature and / or a second feature; The filtering module is specifically used to determine the different user feature as the first feature if the feature weight of the different user feature is greater than the preset feature weight. If there are multiple first features, the multiple first features are combined to obtain the second feature.

10. The resource display device according to claim 7, characterized in that, The feature determination module is specifically used to determine the resource feature as the common resource feature if the proportion of the third user corresponding to the resource feature among the first user corresponding to the resource feature is greater than or equal to a first preset ratio. If the proportion of the third user corresponding to the scene feature among the first user corresponding to the scene feature is greater than or equal to a second preset ratio, the scene feature is determined as the common scene feature.

11. The resource display device according to claim 7, characterized in that, The feature determination module is specifically used to determine the first degree of difference of the common resource feature based on the number of the second user and the number of the third user corresponding to the common resource feature; The second degree of difference of the common scene feature is determined based on the number of the second user and the number of the third user corresponding to the common scene feature.

12. The resource display device according to any one of claims 7 to 11, characterized in that, The output results include interrelated target resource characteristics, target scene characteristics, and purchase probability; the device also includes: The estimation module is configured to, if a preset resource contains the characteristics of the target resource, calculate the product of the price of the preset resource and the purchase probability as the estimated conversion amount corresponding to the characteristics of the target resource; and, If the estimated conversion amount corresponding to the target resource feature is greater than or equal to the first threshold, and the purchase probability is greater than or equal to the second threshold, the preset resource containing the target resource feature is determined as the target resource; The display module is specifically used to display a purchase control for the target resource on the order interface, which includes the target resource characteristics associated with the target scenario characteristics, if the scenario characteristics of the business scenario in which the target user is located match the target scenario characteristics.

13. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps of the method for displaying the resources as described in any one of claims 1 to 6.

14. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the program, it implements the method described in claim 1.

Images