Commodity selection method and device, electronic device, and storage medium

Abstract

This specification provides a product selection method and device, electronic device, and storage medium. The method includes: obtaining a sample set to be trained, wherein the sample set to be trained includes a feature combination consisting of business district features of a sample business district and product features of sample products, and each feature combination is added with a training label, and the training label is used to characterize whether the historical data of the corresponding sample product in the corresponding sample business district meets predefined product selection conditions; based on the sample set to be trained, a constructed neural network model is trained to generate a corresponding product evaluation model, and the product evaluation model is used to predict whether the target product meets the product selection conditions in the target business district.

Description

Technical Field

[0001] This specification relates to the field of data processing technology, and in particular to a product selection method and device, electronic equipment, and storage medium. Background Art

[0002] With the development of the Internet, people are beginning to get used to purchasing goods based on the purchase recommendations provided by shopping platforms. At the same time, major manufacturers and platforms are gradually facing the difficult problem of how to provide products that better meet the needs of users.

[0003] In related technologies, the features of each product are usually extracted and classified, and corresponding product information is provided to users based on the type of product. However, this method does not take the user's own preferences into consideration, resulting in users easily obtaining products that they do not like or are not satisfied with, which in turn leads to a decline in the user's purchasing experience. Summary of the Invention

[0004] In view of this, this specification provides a product selection method and device, electronic device, and storage medium to address the deficiencies in the related art.

[0005] Specifically, this specification is implemented through the following technical solutions:

[0006] According to a first aspect of an embodiment of this specification, a commodity selection method is provided, the method comprising:

[0007] Obtain a sample set to be trained, wherein the sample set to be trained includes a feature combination consisting of business district features of a sample business district and product features of sample products, and each feature combination is added with a training label, wherein the training label is used to indicate whether the historical data of the corresponding sample product in the corresponding sample business district meets the predefined product selection conditions;

[0008] The constructed neural network model is trained based on the sample set to be trained to generate a corresponding product evaluation model, where the product evaluation model is used to predict whether a target product meets the product selection conditions in a target business district.

[0009] Optionally, the process of generating the business district characteristics of any business district includes:

[0010] Aggregate user features of all users in any one of the business districts, and use the generated aggregated user features in the business district as the business district features of the any one of the business districts; or

[0011] The aggregated user features within the business district of any business district and the external business district features from other business districts associated with the business district are integrated, and the generated integrated features are used as the business district features of the business district.

[0012] Optionally, the aggregating user features of all users in any business district includes:

[0013] Dividing the user characteristics of all users in any business district into multiple groups of user characteristics based on preset dimensions;

[0014] Aggregate each group of user features separately to obtain the corresponding initial aggregated user features;

[0015] Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

[0016] Optionally, the preset dimension is at least one of the following: time period, age, gender, and job type.

[0017] Optionally, the user characteristics include user attribute information and / or historical behavior information.

[0018] Optionally, the distance between the other business districts and any of the business districts is not greater than a preset distance.

[0019] Optionally, the process of generating the external business district features of other business districts associated with any business district includes:

[0020] Inputting the adjacency matrix of the any business district and the other business districts, as well as the aggregated user features within the business district of the any business district and the other business districts, into a pre-trained graph convolutional neural network model;

[0021] Determine the external business district features of any business district according to the output results of the graph convolutional neural network model.

[0022] Optionally, the product evaluation model is a neural network model based on an attention mechanism.

[0023] According to a second aspect of the embodiments of this specification, a product selection method is provided, the method comprising:

[0024] Obtaining model input features, where the model input features include a combination of a target business district feature and a target product feature;

[0025] The model input features are input into a pre-trained commodity evaluation model, and it is determined whether the target commodity meets the predefined commodity selection conditions according to the output result of the commodity evaluation model.

[0026] Optionally, also include:

[0027] Aggregate user features of all users in the target business district, and use the generated aggregated user features in the business district as the business district features of the target business district; or

[0028] The aggregated user features within the target business district and the external business district features from other business districts associated with the target business district are fused, and the generated fused features are used as the business district features of the target business district.

[0029] Optionally, aggregating user features of all users in the target business district includes:

[0030] Dividing the user characteristics of all users in the target business district into multiple groups of user characteristics based on preset dimensions;

[0031] Aggregate each group of user features separately to obtain the corresponding initial aggregated user features;

[0032] Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

[0033] Optionally, the preset dimension is at least one of the following: time period, age, gender, and job type.

[0034] Optionally, the user characteristics include user attribute information and / or historical behavior information.

[0035] Optionally, the distance between the other business districts and the target business district is no greater than a preset distance.

[0036] Optionally, also include:

[0037] Inputting the adjacency matrix of the target business district and the other business districts, as well as the aggregated user features within the target business district and the other business districts, into a pre-trained graph convolutional neural network model;

[0038] Determine the external business district features of the target business district according to the output results of the graph convolutional neural network model.

[0039] Optionally, the product evaluation model is a neural network model based on an attention mechanism.

[0040] According to a third aspect of the embodiments of this specification, a commodity selection device is provided, the device comprising:

[0041] A to-be-trained sample set acquisition unit is configured to acquire a to-be-trained sample set, wherein the to-be-trained sample set includes a feature combination consisting of business district features of a sample business district and product features of sample products, and each feature combination is added with a training label, wherein the training label is used to indicate whether the historical data of the corresponding sample product in the corresponding sample business district meets the predefined product selection conditions;

[0042] The model training unit is used to train the constructed neural network model based on the sample set to be trained to generate a corresponding product evaluation model, and the product evaluation model is used to predict whether the target product meets the product selection conditions in the target business district.

[0043] Optionally, the device further includes:

[0044] a business district feature generating unit, configured to aggregate user features of all users in any business district, and use the generated aggregated user features in the business district as the business district feature of the any business district; or

[0045] The aggregated user features within the business district of any business district and the external business district features from other business districts associated with the business district are integrated, and the generated integrated features are used as the business district features of the business district.

[0046] Optionally, the business district feature generating unit is specifically configured to:

[0047] Dividing the user characteristics of all users in any business district into multiple groups of user characteristics based on preset dimensions;

[0048] Aggregate each group of user features separately to obtain the corresponding initial aggregated user features;

[0049] Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

[0050] Optionally, the preset dimension is at least one of the following: time period, age, gender, and job type.

[0051] Optionally, the user characteristics include user attribute information and / or historical behavior information.

[0052] Optionally, the distance between the other business districts and any of the business districts is not greater than a preset distance.

[0053] Optionally, the device further includes:

[0054] An external business district feature generation unit is configured to input the adjacency matrix of the any business district and the other business districts, as well as the aggregated user features within the business district of the any business district and the other business districts, into a pre-trained graph convolutional neural network model;

[0055] Determine the external business district features of any business district according to the output results of the graph convolutional neural network model.

[0056] Optionally, the product evaluation model is a neural network model based on an attention mechanism.

[0057] According to a fourth aspect of the embodiments of this specification, a commodity selection device is provided, the device comprising:

[0058] A model input feature acquisition unit, configured to acquire model input features, wherein the model input features include a combination of a target business district feature and a target product feature;

[0059] The product selection unit is used to input the model input features into a pre-trained product evaluation model, and determine whether the target product meets the predefined product selection conditions according to the output result of the product evaluation model.

[0060] Optionally, the device further includes:

[0061] a target business district feature generating unit, configured to aggregate user features of all users in the target business district, and use the generated aggregated user features in the business district as the business district feature of the target business district; or

[0062] The aggregated user features within the target business district and the external business district features from other business districts associated with the target business district are fused, and the generated fused features are used as the business district features of the target business district.

[0063] Optionally, the target business district feature generation unit is specifically configured to:

[0064] Dividing the user characteristics of all users in the target business district into multiple groups of user characteristics based on preset dimensions;

[0065] Aggregate each group of user features separately to obtain the corresponding initial aggregated user features;

[0066] Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

[0067] Optionally, the preset dimension is at least one of the following: time period, age, gender, and job type.

[0068] Optionally, the user characteristics include user attribute information and / or historical behavior information.

[0069] Optionally, the distance between the other business districts and the target business district is no greater than a preset distance.

[0070] Optionally, the device further includes:

[0071] A target external business district feature generation unit is configured to input the adjacency matrix of the target business district and the other business districts, as well as the aggregated user features within the target business district and the other business districts, into a pre-trained graph convolutional neural network model;

[0072] Determine the external business district features of the target business district according to the output results of the graph convolutional neural network model.

[0073] Optionally, the product evaluation model is a neural network model based on an attention mechanism.

[0074] According to a fifth aspect of the embodiments of this specification, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the steps of the method described in the first and second aspects are implemented.

[0075] According to the sixth aspect of the embodiments of this specification, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method described in the first and second aspects are implemented.

[0076] In the technical solution presented in this specification, a neural network model is trained using a combination of features consisting of both business district and product characteristics. This neural network model can then be used to determine whether a target product meets the corresponding product selection criteria within a target business district. In other words, this technical solution combines product and business district characteristics within the neural network model, ensuring a high degree of match between selected products and user preferences within the business district, thereby enhancing the user purchasing experience.

[0077] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0079] Figure 1 This is a flowchart of a commodity selection method shown in an exemplary embodiment of this specification;

[0080] Figure 2 This is a flowchart of another commodity selection method shown in an exemplary embodiment of this specification;

[0081] Figure 3 is a schematic diagram of a commodity selection method shown in an exemplary embodiment of this specification;

[0082] Figure 4 is a schematic diagram of another commodity selection method shown in an exemplary embodiment of this specification;

[0083] Figure 5This is a schematic diagram of a business district relationship shown in an exemplary embodiment of this specification;

[0084] Figure 6 is a schematic structural diagram of an electronic device shown in an exemplary embodiment of this specification;

[0085] Figure 7 This is a schematic structural diagram of a commodity selection device shown in an exemplary embodiment of this specification;

[0086] Figure 8 It is a structural diagram of another commodity selection device shown in an exemplary embodiment of this specification. DETAILED DESCRIPTION

[0087] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with this specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of this specification, as detailed in the appended claims.

[0088] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. As used in this specification and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0089] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, such information should not be limited to these terms. These terms are merely used to distinguish information of the same type from one another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information without departing from the scope of this specification. Depending on the context, the term "if" as used herein may be interpreted as "when," "when," or "in response to determining."

[0090] The following combination Figure 1 The illustrated embodiments illustrate the technical solutions of this specification. Figure 1 This is a flowchart of a commodity selection method shown in an exemplary embodiment of this specification. Figure 1 As shown, the method may include the following steps:

[0091] S101, obtaining a sample set to be trained, wherein the sample set to be trained includes a feature combination consisting of business district features of a sample business district and product features of sample products, and each feature combination is added with a training label, and the training label is used to indicate whether the historical data of the corresponding sample product in the corresponding sample business district meets the predefined product selection conditions.

[0092] The above-mentioned sample set to be trained serves as sample data for training the product evaluation model below, and may include one or more business district features of the sample business district and product features of the sample products. The business district features and product features may be combined in pairs, and corresponding training labels may be added to the feature combination based on the historical data of the sample products of the above-mentioned feature combination in the corresponding sample business district, so that the training objective of the above-mentioned product evaluation model can be associated with the performance ability of any product in any business district.

[0093] Since users are the final buyers of goods, the business district characteristics of any of the above business districts can be regarded as user characteristics within the business district. At the same time, the above business district characteristics can be designed in a variety of ways to generate them according to actual conditions, and this manual does not limit this.

[0094] In one embodiment, the user features of all users in any of the above-mentioned business circles can be aggregated, and the generated aggregated user features in the business circle can be used as the business circle features of any of the above-mentioned business circles. Aggregation is a relatively mature technical means in this field, which is used to select, analyze, and classify multiple data of the same dimension. For example: there are 10 users and they are named as users 1 to 10 respectively, and the user features of the above 10 users are X1 to X10 respectively. X0 can be obtained by aggregating X1 to X10, and X0 can be used to represent the overall user features of users 1 to 10. In addition, this specification does not limit the specific method of implementing aggregation. For example: the user features of all the above-mentioned users can be converted into an overall user feature through a preset aggregation function or a pre-trained neural network model.

[0095] In another embodiment, the aggregated user features within the business district of any of the above-mentioned business districts and the external business district features from other business districts associated with any of the above-mentioned business districts can be fused, and the generated fused features can be used as the business district features of any of the above-mentioned business districts. Among them, as for the aggregated user features within the business district of any of the above-mentioned business districts, the generation method is basically the same as that of the previous embodiment, and this specification will not go into details here. As for the external business district features of other business districts associated with any of the above-mentioned business districts, the external business district features do not have a direct relationship with the user features of any of the above-mentioned business districts. Since the business district is a collection of geographical dimensions formed by artificial divisions and including different merchants, the geographical meaning given to it naturally makes different business districts in the same area have certain similarities. Therefore, the external business district features of any of the above-mentioned business districts can be reasonably inferred and predicted based on the business district features of other business districts. In addition, similar to aggregation, fusion is also a relatively mature technical means in this field. The difference between the two is that fusion specifically refers to the selection, analysis, and classification of multiple data of different dimensions. Furthermore, since the above-mentioned external business district characteristics can be used as consistent factors to determine the business district characteristics of any of the above-mentioned business districts, compared with the solution in the previous embodiment in which the business district characteristics of any of the above-mentioned business districts are determined only based on the aggregated user characteristics within the business district, this embodiment determines the business district characteristics of any of the above-mentioned business districts based on the two dimensions of "aggregated user characteristics within the business district" + "external business district characteristics", so that the final business district characteristics are closer to the actual situation, and also promote the improvement of the prediction accuracy and credibility of the product evaluation model below.

[0096] When the user base of a business district is large, all users of the business district can be divided according to preset dimensions, thereby obtaining multiple targeted user groups and their corresponding user characteristics.

[0097] In one embodiment, the user features of all users in any of the above-mentioned business districts can be divided into multiple groups of user features based on preset dimensions, and each group of user features can be aggregated separately to obtain the corresponding initial aggregated user features, and then all the initial aggregated user features can be aggregated to obtain the aggregated user features in the above-mentioned business district. Among them, the above-mentioned preset dimensions can include at least one of the following dimensions: time period, age, gender, and job type, which are not limited in this specification. Since the users are first classified once in this implementation, the above-mentioned initial aggregated user features can further highlight the user features of a part of the users in any of the above-mentioned business districts. Therefore, the aggregated user features in the business district after aggregating all the initial aggregated user features are closer to the actual situation than the aggregated user features in the business district obtained by directly aggregating all user features, and are beneficial to the prediction effect of the product evaluation model below.

[0098] The above-mentioned user characteristics may include user attribute information and / or historical behavior information. Among them, the above-mentioned user attribute information corresponds to the above-mentioned preset dimensions, that is, it represents relatively fixed personal information of the user such as age, gender, etc. The above-mentioned historical behavior information can be the user's historical statistical behavior, for example: the number of purchases and orders made by the user in the past 10 days, or the frequency of purchasing the same product, etc. It should be emphasized that the collection and uploading operations of the above-mentioned user characteristics are based on the provisions of the relevant privacy policy and are performed with the authorization and permission of the user. If the user refuses to authorize and permit the above-mentioned actions, the user characteristics of the user will not be collected.

[0099] In addition, a preset distance can be set so that the distance between the above-mentioned other business districts and any of the above-mentioned business districts has to be greater than the preset distance. For example, as mentioned above, different business districts in the same area naturally have certain similarities. Therefore, limiting the distance between the above-mentioned other business districts and any of the above-mentioned business districts can ensure the accuracy of the above-mentioned external business district characteristics, and the preset distance is negatively correlated with the accuracy of the external business district characteristics within a certain range.

[0100] For the above-mentioned external business district characteristics, a suitable generation method can be selected according to actual needs.

[0101] In one embodiment, the aggregated user characteristics within each of the aforementioned business districts and the aforementioned other business districts may be input into a preset formula to obtain the external business district characteristics of the aforementioned business district. The preset formula may be set based on an average of the aggregated user characteristics within each of the aforementioned other business districts, or may be weighted based on a preset weight to calculate the external business district characteristics of the aforementioned business district.

[0102] In another embodiment, the adjacency matrix between any of the aforementioned business districts and the aforementioned other business districts, as well as the aggregated user features within each of the aforementioned business districts and the aforementioned other business districts, can be input into a pre-trained graph convolutional neural network model. The external business district features of any of the aforementioned business districts can then be determined based on the output of the graph convolutional neural network model. Compared to the previous embodiment, this embodiment leverages the autonomous learning properties of neural network technology to effectively improve the efficiency of calculating external business district features.

[0103] S102: Training the constructed neural network model based on the sample set to be trained to generate a corresponding product evaluation model, where the product evaluation model is used to predict whether the target product meets the product selection condition in the target business district.

[0104] After training on the aforementioned sample set, the constructed neural network model can generate a corresponding product evaluation model. When a target product and a target business district are input, the product evaluation model can predict whether the target product meets the aforementioned product selection criteria within the target business district, thereby helping users within the target business district filter out unsuitable and select suitable target products. The product evaluation model can be a neural network model based on an attention mechanism, or a neural network model of any other type, and this specification does not limit this.

[0105] The above-mentioned product selection conditions correspond to the above-mentioned training labels and can be set according to actual needs. For example, when it is necessary to provide the "most popular" products to users in the target business district, the above-mentioned product selection conditions can be set to the product exposure click rate. For example, when it is necessary to provide "high-reputation" products to users in the target business district, the above-mentioned product selection conditions can be set to the product praise rate. This manual does not limit this.

[0106] The following combination Figure 2 The illustrated embodiments illustrate the technical solutions of this specification. Figure 2 This is a flowchart of another commodity selection method shown in an exemplary embodiment of this specification. Figure 2 As shown, the method may include the following steps:

[0107] S201 , obtaining model input features, where the model input features include a combination of a target business district feature and a target product feature.

[0108] The above-mentioned model input features serve as input parameters for prediction by the product evaluation model below, and may include a combination of the business district features of the target business district and the product features of the target product. This specification does not limit the total number of business district features and product features in the above-mentioned model input features, as well as the specific combination method. According to actual conditions, it may be stipulated that the number of the above-mentioned product features is not less than 500, or that the business district features are only combined with 50% of the product features. At the same time, this specification does not limit the relationship between the target business district and the above-mentioned sample business districts, or the target product and the above-mentioned sample products. The above-mentioned target business district may be a newly defined new business district, or any of the above-mentioned sample business districts, and the above-mentioned target business district may be a newly launched product, or any of the above-mentioned sample products.

[0109] Since users are the final buyers of goods, the business district characteristics of the above-mentioned target business district can be regarded as user characteristics within the business district. At the same time, the above-mentioned business district characteristics can be designed in a variety of ways to generate according to actual conditions, and this manual does not limit this.

[0110] In one embodiment, the user features of all users in the target business district can be aggregated, and the generated aggregated user features in the business district can be used as the business district features of the target business district. Aggregation is a relatively mature technical means in this field, which is used to select, analyze and classify multiple data of the same dimension. For example: there are 10 users and they are named as users 1 to 10 respectively, and the user features of the above 10 users are X1 to X10 respectively. X0 can be obtained by aggregating X1 to X10, and X0 can be used to represent the overall user features of users 1 to 10. In addition, this specification does not limit the specific method of implementing aggregation. For example: the user features of all the above users can be converted into an overall user feature through a preset aggregation function or a pre-trained neural network model.

[0111] In another embodiment, the aggregated user features within the target business district and the external business district features from other business districts associated with the target business district can be fused, and the generated fused features can be used as the business district features of the target business district. Among them, for the aggregated user features within the target business district, the generation method is basically the same as that of the previous embodiment, and this specification will not go into details here. As for the external business district features of other business districts associated with the target business district, there is no direct relationship between the external business district features and the user features of the target business district. Since the business district is a collection of geographical dimensions formed by artificial divisions and including different businesses, the geographical meaning given to it naturally makes different business districts in the same area have certain similarities. Therefore, the external business district features of the target business district can be reasonably inferred and predicted based on the business district features of other business districts. In addition, similar to aggregation, fusion is also a relatively mature technical means in this field. The difference between the two is that fusion specifically refers to the selection, analysis, and classification of multiple data of different dimensions. Furthermore, since the above-mentioned external business district characteristics can be used as the consistent factors for determining the business district characteristics of the above-mentioned target business district, compared with the scheme in the previous embodiment in which the business district characteristics of the above-mentioned target business district are determined only based on the aggregated user characteristics within the business district, this embodiment determines the business district characteristics of the above-mentioned target business district based on the two dimensions of "aggregated user characteristics within the business district" + "external business district characteristics", so that the final business district characteristics are closer to the actual situation, and also promote the improvement of the prediction accuracy and credibility of the product evaluation model below.

[0112] When the user base of a business district is large, all users of the business district can be divided according to preset dimensions, thereby obtaining multiple targeted user groups and their corresponding user characteristics.

[0113] In one embodiment, the user features of all users in the above-mentioned target business district can be divided into multiple groups of user features based on preset dimensions, and each group of user features can be aggregated separately to obtain the corresponding initial aggregated user features, and then all the initial aggregated user features can be aggregated to obtain the aggregated user features in the above-mentioned business district. Among them, the above-mentioned preset dimensions can include at least one of the following dimensions: time period, age, gender, and job type, which are not limited in this specification. Since the users are first classified once in this implementation, the above-mentioned initial aggregated user features can further highlight the user features of a part of the users in the above-mentioned target business district. Therefore, the aggregated user features in the business district after aggregating all the initial aggregated user features are closer to the actual situation than the aggregated user features in the business district obtained by directly aggregating all user features, and are beneficial to the prediction effect of the product evaluation model below.

[0114] The above-mentioned user characteristics may include user attribute information and / or historical behavior information. Among them, the above-mentioned user attribute information corresponds to the above-mentioned preset dimensions, that is, it represents relatively fixed personal information of the user such as age, gender, etc. The above-mentioned historical behavior information can be the user's historical statistical behavior, for example: the number of purchases and orders made by the user in the past 10 days, or the frequency of purchasing the same product, etc. It should be emphasized that the collection and uploading operations of the above-mentioned user characteristics are based on the provisions of the relevant privacy policy and are performed with the authorization and permission of the user. If the user refuses to authorize and permit the above-mentioned behavior, the user characteristics of the user will be ignored.

[0115] In addition, a preset distance can be set so that the distance between the above-mentioned other business districts and the above-mentioned target business district has to be greater than the preset distance. For example, as mentioned above, different business districts in the same area naturally have certain similarities. Therefore, limiting the distance between the above-mentioned other business districts and the above-mentioned target business district can ensure the accuracy of the above-mentioned external business district characteristics, and the preset distance is negatively correlated with the accuracy of the external business district characteristics within a certain range.

[0116] For the above-mentioned external business district characteristics, a suitable generation method can be selected according to actual needs.

[0117] In one embodiment, the aggregated user characteristics within the target business district and the other business districts may be input into a preset formula to obtain the external business district characteristics of the target business district. The preset formula may be based on an average of the aggregated user characteristics within the other business districts, or may be weighted based on the aggregated user characteristics within the other business districts according to preset weights to determine the external business district characteristics of the target business district.

[0118] In another embodiment, the adjacency matrix of the target business district and the other business districts, as well as the aggregated user features within each of the target business district and the other business districts, can be input into a pre-trained graph convolutional neural network model. The external business district features of the target business district can then be determined based on the output of the graph convolutional neural network model. Compared to the previous embodiment, this embodiment leverages the autonomous learning properties of neural network technology to effectively improve the efficiency of calculating external business district features.

[0119] S202: Input the model input features into a pre-trained product evaluation model, and determine whether the target product meets predefined product selection conditions based on an output result of the product evaluation model.

[0120] The output result of the above-mentioned product evaluation model can be a specific number. When the number exceeds a preset threshold, it is determined that the above-mentioned target product meets the above-mentioned predefined product selection conditions. For example: there is a preset threshold of 70% as a percentage value, two target products 1 and 2, and a target business district S. Then, when the output results of "target product 1-target business district S" and "target product 2-target business district S" are 65% and 80% respectively after the above-mentioned product evaluation model, it can be predicted that target product 1 is a product in the target business district S that does not meet the above-mentioned predefined product selection conditions, and target product 2 is a product in the target business district S that meets the above-mentioned predefined product selection conditions. In addition, the above-mentioned product evaluation model can be a neural network model based on the attention mechanism, or a neural network model of various types, which is not limited in this specification.

[0121] The above-mentioned product selection conditions correspond to the above-mentioned training labels and can be set according to actual needs. For example, when it is necessary to provide the "most popular" products to users in the target business district, the above-mentioned product selection conditions can be set to the product exposure click rate. For example, when it is necessary to provide "high-reputation" products to users in the target business district, the above-mentioned product selection conditions can be set to the product praise rate. This manual does not limit this.

[0122] It can be seen from the above embodiments that in this specification, firstly, by combining the business district characteristics and the product characteristics into the neural network model, the neural network model, that is, the trained product evaluation model, can be used to predict whether the target product meets the product selection conditions corresponding to the target business district, thereby improving the satisfaction of users in the business district with the products they browse, and further improving the user shopping experience. Secondly, in this specification, user characteristics can be used as elements to generate business district characteristics, so that the results output by the above product evaluation model are more in line with the actual needs of users. At the same time, the method of classifying and re-aggregating user characteristics ensures that different types of user characteristics can be fully considered in the aggregation process. In addition, the combination of aggregated user characteristics within the business district and external business district characteristics can further improve the effectiveness of the business district characteristics of the corresponding business district.

[0123] The following combination Figure 3 The illustrated embodiments illustrate the technical solutions of this specification. Figure 3 This is a schematic diagram of a commodity selection method shown in an exemplary embodiment of this specification. Figure 3 As shown, the method can be divided into the following steps:

[0124] S301, define a sample business district.

[0125] Define a city with several sample business districts 1, 2, 3...N, and each sample business district contains several businesses.

[0126] S302a, obtaining user characteristics.

[0127] Taking the sample business district N as an example, assuming that there are 10 merchants engaged in the catering industry and 100 users in the sample business district N, all of the above users have made consumption behaviors in at least one of the above 10 merchants. With the permission of each user, user characteristics X1~100 of the above 100 users can be formed based on the age and gender of each user.

[0128] S303a, obtaining aggregated user features within the business district.

[0129] Based on the historical purchasing behavior of the 100 users, we can use behaviors at different time periods (e.g., breakfast, lunch, afternoon tea, dinner, and late-night snacks) to represent user demand within sample business district N. Assume that 50 of the 100 users belong to the group that needs to buy breakfast, 30 belong to the group that needs to buy lunch, and the final 10 belong to the group that needs to buy dinner. By aggregating the user features of these three user groups, namely, aggregating X1-X50, X51-X80, and X81-X100, we obtain three sets of initial aggregated user features: A, B, and C. These three sets of initial aggregated user features correspond to the user groups for breakfast, lunch, and dinner, respectively. We then aggregate the initial aggregated user features A, B, and C to obtain the aggregated user feature Y1 for sample business district N.

[0130] S302b, obtaining a correlation graph of other business districts.

[0131] Assume that there is a preset distance of 1 km, and the sample business districts 1 to N-1 are no more than the preset distance from the sample business district N. The adjacency matrix between the sample business district N and the sample business districts 1, 2, and 3 can be constructed by constructing a relationship graph in the above city. Figure 5 The illustrated embodiments illustrate the technical solutions of this specification. Figure 5 This is a schematic diagram of a business district relationship shown in an exemplary embodiment of this specification, where sample business district N is only connected to sample business districts 1, 2, and 3. The corresponding adjacency matrix can be obtained as follows:

[0132] Business District N Business District 1 Business District 2 Business District 3 Business District N 0 1 1 1 Business District 1 1 0 1 1 Business District 2 1 1 0 0 Business District 3 1 1 0 0

[0133] S303b, obtaining external business district features.

[0134] The adjacency matrix and the aggregated user features within each of the sample business districts 1, 2, 3, and N are input into a pre-trained graph convolutional neural network model, and the external business district feature Y2 of business district N is determined based on the output results of the graph convolutional neural network model.

[0135] S304: Obtain the business district characteristics of the sample business district.

[0136] The aggregated user features Y1 within the business district and the external business district features Y2 are fused based on the neural network of the attention mechanism to obtain the business district features N of the sample business district N.

[0137] S302c, obtaining product features of the sample product.

[0138] Assuming that there are commodities 1, 2, 3, ..., M, commodity features 1, 2, 3, ..., M of each commodity can be determined based on commodity type, price, and other aspects.

[0139] S305: training a product evaluation model.

[0140] Combine business district feature N with product features 1-M, and determine the corresponding training labels based on the click-through rates of sample products 1-M in sample business district N. This is then input into the constructed neural network model to train it. Repeat this process until all combinations of sample business districts and products are used to train the neural network model, resulting in a trained product evaluation model.

[0141] The following combination Figure 4 The illustrated embodiments illustrate the technical solutions of this specification. Figure 4 This is a schematic diagram of a commodity selection method shown in an exemplary embodiment of this specification. Figure 4 As shown, the method can be divided into the following steps:

[0142] S401, define the target business district.

[0143] Redefine a geographical area within a certain range in any city as target business district N+1, and the target business district N+1 is different from any of the sample business districts.

[0144] S402a, obtaining user characteristics.

[0145] Obtain user characteristics of all users in the target business district. For specific operations, please refer to S302a.

[0146] S403a, obtaining aggregated user features within the business district.

[0147] The user features of all users in the target business district are combined according to time periods to obtain multiple groups of aggregated user features within the business district. For specific operations, please refer to S303a.

[0148] S402b: Obtain a correlation graph of other business districts.

[0149] Assuming that the distance between target business district N+1 and business districts N+2 and N+3 is not greater than a preset distance based on the city construction relationship diagram of the city where the target business district is located, an adjacency matrix between target business district N+1 and business districts N+2 and N+3 can be constructed. The specific operation is described in S302b.

[0150] S403b, obtaining external business district features.

[0151] The adjacency matrix between the target business district N+1 and business districts N+2 and N+3, as well as the aggregated user features within the target business districts N+1, N+2, and N+3, are input into the pre-trained graph convolutional neural network model, and the external business district features of business district N+1 are determined based on the output results of the graph convolutional neural network model.

[0152] S404: Acquire the business district characteristics of the target business district.

[0153] The aggregated user features within the target business district N+1 and the external business district features are fused based on the neural network of the attention mechanism to obtain the business district feature N+1 of the target business district N.

[0154] S402c, obtaining product features of the target product.

[0155] Assume that there is a newly launched target product M+1, and product feature M+1 of the target product M+1 can be determined based on product type, price, and other aspects.

[0156] S405: Output the model output result.

[0157] To predict whether target product M+1 is a high-quality product in target business district N+1 (assuming high-quality products are products with high exposure and click-through rates), the above operation can be used to obtain the business district feature YN+1 of target business district N+1 and the product feature ZM+1 of target product M+1, and input these features into the aforementioned product evaluation model. Assuming that the product evaluation model has a preset threshold of 80% and the obtained output result is 90%, it can be predicted whether target product M+1 is a high-quality product in target business district N+1.

[0158] Figure 6 This is a schematic structural diagram of an electronic device in an exemplary embodiment. Figure 6 At the hardware level, the electronic device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other necessary hardware. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a product selection device at the logical level. Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0159] Corresponding to the aforementioned embodiments of the commodity selection method, this specification also provides two embodiments of a commodity selection device.

[0160] Please refer to Figure 7 , Figure 7 FIG. 1 is a schematic diagram showing the structure of a commodity selection device according to an exemplary embodiment. Figure 7 As shown, in a software implementation, the device may include:

[0161] The training sample set acquisition unit 701 is used to acquire a training sample set, wherein the training sample set includes a feature combination consisting of the business district features of the sample business district and the product features of the sample products, and each feature combination is added with a training label, wherein the training label is used to indicate whether the historical data of the corresponding sample product in the corresponding sample business district meets the predefined product selection conditions;

[0162] The model training unit 702 is used to train the constructed neural network model based on the sample set to be trained to generate a corresponding product evaluation model, which is used to predict whether the target product meets the product selection conditions in the target business district.

[0163] Optionally, the device further includes:

[0164] The business district feature generating unit 703 is configured to aggregate user features of all users in any business district and use the generated aggregated user features in the business district as the business district feature of the any business district; or

[0165] The aggregated user features within the business district of any business district and the external business district features from other business districts associated with the business district are integrated, and the generated integrated features are used as the business district features of the business district.

[0166] Optionally, the business district feature generating unit 703 is specifically configured to:

[0167] Dividing the user characteristics of all users in any business district into multiple groups of user characteristics based on preset dimensions;

[0168] Aggregate each group of user features separately to obtain the corresponding initial aggregated user features;

[0169] Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

[0170] Optionally, the preset dimension is at least one of the following: time period, age, gender, and job type.

[0171] Optionally, the user characteristics include user attribute information and / or historical behavior information.

[0172] Optionally, the distance between the other business districts and any of the business districts is not greater than a preset distance.

[0173] Optionally, the device further includes:

[0174] The external business district feature generation unit 704 is configured to input the adjacency matrix between the one business district and the other business districts, and the aggregated user features within the business districts of the one business district and the other business districts into a pre-trained graph convolutional neural network model.

[0175] Determine the external business district features of any business district according to the output results of the graph convolutional neural network model.

[0176] Optionally, the product evaluation model is a neural network model based on an attention mechanism.

[0177] Please refer to Figure 8 , Figure 8 FIG. 1 is a schematic diagram showing the structure of another commodity selection device according to an exemplary embodiment. Figure 8 As shown, in a software implementation, the device may include:

[0178] The model input feature acquisition unit 801 is used to acquire model input features, wherein the model input features include a combination of the target business district features and the target product features;

[0179] The product selection unit 802 is configured to input the model input features into a pre-trained product evaluation model, and determine whether the target product meets a predefined product selection condition based on an output result of the product evaluation model.

[0180] Optionally, the device further includes:

[0181] The target business district feature generating unit 803 is configured to aggregate user features of all users in the target business district and use the generated aggregated user features in the business district as the business district feature of the target business district; or

[0182] The aggregated user features within the target business district and the external business district features from other business districts associated with the target business district are fused, and the generated fused features are used as the business district features of the target business district.

[0183] Optionally, the target business district feature generating unit 803 is specifically configured to:

[0184] Dividing the user characteristics of all users in the target business district into multiple groups of user characteristics based on preset dimensions;

[0185] Aggregate each group of user features separately to obtain the corresponding initial aggregated user features;

[0186] Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

[0187] Optionally, the preset dimension is at least one of the following: time period, age, gender, and job type.

[0188] Optionally, the user characteristics include user attribute information and / or historical behavior information.

[0189] Optionally, the distance between the other business districts and the target business district is no greater than a preset distance.

[0190] Optionally, the device further includes:

[0191] The target external business district feature generation unit 804 is configured to input the adjacency matrix of the target business district and the other business districts, as well as the aggregated user features within the target business district and the other business districts, into a pre-trained graph convolutional neural network model.

[0192] Determine the external business district features of the target business district according to the output results of the graph convolutional neural network model.

[0193] Optionally, the product evaluation model is a neural network model based on an attention mechanism.

[0194] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0195] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this specification. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0196] Embodiments of the subject matter and functional operations described in this specification may be implemented in the following: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. Embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier to be executed by a data processing device or to control the operation of the data processing device. Alternatively or additionally, the program instructions may be encoded on an artificially generated propagation signal, such as a machine-generated electrical, optical, or electromagnetic signal, which is generated to encode information and transmit it to a suitable receiver device for execution by the data processing device. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.

[0197] The processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform the corresponding functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can be implemented as, special purpose logic circuitry, such as an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).

[0198] Computers suitable for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit will receive instructions and data from a read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as magnetic disks, magneto-optical disks, or optical disks, or the computer will be operably coupled to such mass storage devices to receive data from them or to transmit data to them, or both. However, a computer does not necessarily have such devices. In addition, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0199] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and storage devices, including, for example, semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD ROM and DVD-ROM disks. The processor and memory can be supplemented by, or incorporated in, special purpose logic circuitry.

[0200] Although this specification includes many specific implementation details, these should not be interpreted as limiting the scope of any invention or the scope of protection claimed, but are mainly used to describe the features of specific embodiments of specific inventions. Certain features described in multiple embodiments within this specification may also be implemented in combination in a single embodiment. On the other hand, the various features described in a single embodiment may also be implemented separately in multiple embodiments or in any suitable sub-combination. In addition, although features may work in certain combinations as described above and even initially claimed as such, one or more features from the claimed combination may be removed from the combination in some cases, and the claimed combination may point to a sub-combination or a variation of the sub-combination.

[0201] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that these operations be performed in the particular order shown or performed sequentially, or that all illustrated operations be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

[0202] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the particular order shown or sequential sequence to achieve the desired results. In some implementations, multitasking and parallel processing may be advantageous.

[0203] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.

Claims

1. A commodity selection method, characterized in that: The method comprises: Obtain a sample set to be trained, wherein the sample set to be trained includes a feature combination consisting of business district features of a sample business district and product features of sample products, and each feature combination is added with a training label, wherein the training label is used to indicate whether the historical data of the corresponding sample product in the corresponding sample business district meets the predefined product selection conditions; The constructed neural network model is trained based on the training sample set to generate a corresponding product evaluation model, wherein the training objective of the product evaluation model is to learn the performance of any product in any business district, and the product evaluation model is used to predict whether the target product meets the product selection conditions in the target business district. The process of generating the business district characteristics of any business district includes: Aggregate user features of all users in any one of the business districts, and use the generated aggregated user features in the business district as the business district features of the any one of the business districts; or The aggregated user features within the business district of any business district and the external business district features from other business districts associated with the business district are integrated, and the generated integrated features are used as the business district features of the business district.

2. The method according to claim 1, characterized in that The user characteristics of all users in any business circle are aggregated, including: Dividing the user characteristics of all users in any business district into multiple groups of user characteristics based on preset dimensions; Aggregate each group of user features separately to obtain the corresponding initial aggregated user features; Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

3. The method according to claim 2, characterized in that The preset dimension is at least one of the following: time period, age, gender, and job type.

4. The method according to claim 1, wherein The user characteristics include user attribute information and / or historical behavior information.

5. The method according to claim 1, characterized in that The distance between the other business districts and any of the business districts is no greater than a preset distance.

6. The method according to claim 1, characterized in that The process of generating the external business district features of other business districts associated with any business district includes: Inputting the adjacency matrix of the any business district and the other business districts, as well as the aggregated user features within the business district of the any business district and the other business districts, into a pre-trained graph convolutional neural network model; Determine the external business district features of any business district according to the output results of the graph convolutional neural network model.

7. The method according to claim 1, characterized in that The product evaluation model is a neural network model based on the attention mechanism.

8. A commodity selection method, characterized in that: The method comprises: Obtaining model input features, wherein the model input features include a combination of business district features of a target business district and product features of a target product; wherein the business district features of the target business district are determined based on the following method: aggregating user features of all users in the target business district and using the generated aggregated user features in the business district as the business district features of the target business district; or, Fusing aggregated user features within the target business district with external business district features from other business districts associated with the target business district, and using the generated fused features as the business district features of the target business district; The model input features are input into a pre-trained product evaluation model, and whether the target product meets the predefined product selection conditions is determined based on the output results of the product evaluation model. In the process of training the product evaluation model, the training goal of the product evaluation model is to learn the performance ability of any product in any business district.

9. The method according to claim 8, characterized in that The user characteristics of all users in the target business district are aggregated, including: Dividing the user characteristics of all users in the target business district into multiple groups of user characteristics based on preset dimensions; Aggregate each group of user features separately to obtain the corresponding initial aggregated user features; Aggregate all initial aggregated user features to obtain aggregated user features within the business district.

10. The method according to claim 9, characterized in that The preset dimension is at least one of the following: time period, age, gender, and job type.

11. The method according to claim 8, characterized in that The user characteristics include user attribute information and / or historical behavior information.

12. The method according to claim 8, characterized in that The distance between the other business districts and the target business district is no greater than a preset distance.

13. The method according to claim 8, characterized in that Also includes: Inputting the adjacency matrix of the target business district and the other business districts, as well as the aggregated user features within the target business district and the other business districts, into a pre-trained graph convolutional neural network model; Determine the external business district features of the target business district according to the output results of the graph convolutional neural network model.

14. The method according to claim 8, characterized in that The product evaluation model is a neural network model based on the attention mechanism.

15. A commodity selection device, characterized in that: The device comprises: A to-be-trained sample set acquisition unit is configured to acquire a to-be-trained sample set, wherein the to-be-trained sample set includes a feature combination consisting of business district features of a sample business district and product features of sample products, and each feature combination is added with a training label, wherein the training label is used to indicate whether the historical data of the corresponding sample product in the corresponding sample business district meets the predefined product selection conditions; A model training unit is configured to train the constructed neural network model based on the training sample set to generate a corresponding product evaluation model, wherein the training objective of the product evaluation model is to learn the performance of any product in any business district, and the product evaluation model is used to predict whether the target product meets the product selection conditions in the target business district; wherein the process of generating the business district characteristics of any business district includes: Aggregate user features of all users in any one of the business districts, and use the generated aggregated user features in the business district as the business district features of the any one of the business districts; or The aggregated user features within the business district of any business district and the external business district features from other business districts associated with the business district are integrated, and the generated integrated features are used as the business district features of the business district.

16. A commodity selection device, characterized in that: The device comprises: A model input feature acquisition unit is used to acquire model input features, wherein the model input features include a combination of the business district features of the target business district and the product features of the target product; wherein the business district features of the target business district are determined based on the following method: aggregating user features of all users in the target business district and using the generated aggregated user features in the business district as the business district features of the target business district; or Fusing aggregated user features within the target business district with external business district features from other business districts associated with the target business district, and using the generated fused features as the business district features of the target business district; A product selection unit is used to input the model input features into a pre-trained product evaluation model, and determine whether the target product meets the predefined product selection conditions based on the output results of the product evaluation model. In the process of training the product evaluation model, the training goal of the product evaluation model is to learn the performance ability of any product in any business district.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 14 are implemented.

18. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 14 are implemented.

Images