Prediction Method and Device for Potential Users

Abstract

An embodiment of the present application provides a method and device for predicting potential users, which can be applied to the field of big data analysis. The method includes: determining a first influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first historical behavior data of multiple users for each commodity category within a first time period; determining a second influence weight of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted according to the second historical behavior data of multiple users for each commodity category within a second time period; determining a target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first influence weight and the second influence weight, and determining potential users corresponding to the commodity category to be predicted among multiple users according to the target influence weight. The embodiment of the present application can improve the accuracy of the potential user prediction result.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of big data analysis, and in particular, to a method and device for predicting potential users. Background Art

[0002] In the field of e-commerce, during the process of users selecting and purchasing goods on an e-commerce platform, various operation behaviors such as browsing, favoriting, adding to cart, and purchasing usually occur. By analyzing these operation behaviors, the shopping habits, preferences, or demands of users can be predicted, and then based on the prediction results, goods can be recommended to potential users in a targeted manner, which can provide a better shopping experience for users.

[0003] In traditional technologies, generally, based on the historical operation behaviors of users, traditional machine learning methods such as statistics and regression analysis are used to mine and predict potential users who may purchase a certain commodity.

[0004] However, traditional technologies only consider the overall historical operation behaviors of users and do not distinguish the operation behaviors of users in terms of time and operation sequence, resulting in inaccurate prediction results. Summary of the Invention

[0005] The embodiments of the present application provide a method and device for predicting potential users, which can effectively improve the accuracy of the prediction results of potential users.

[0006] In a first aspect, the embodiments of the present application provide a method for predicting potential users, and the method includes:

[0007] Determine a first influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first historical behavior data of multiple users for each commodity category in a first time period;

[0008] Determine a second influence weight of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted according to the second historical behavior data of the multiple users for each commodity category in a second time period;

[0009] Determine a target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first influence weight and the second influence weight, and determine potential users corresponding to the commodity category to be predicted among the multiple users according to the target influence weight.

[0010] In a feasible implementation manner, the determining a first influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted includes:

[0011] Aggregate various user behaviors corresponding to each commodity category in the first historical behavior data to obtain an aggregation result, where the aggregation result includes a two-dimensional matrix. The first dimension of the two-dimensional matrix is the user identifier of each user, and the second dimension is various user behaviors corresponding to each commodity category;

[0012] Input the aggregation result into a preset Deep Neural Networks (DNN) model for training to determine the first influence weight.

[0013] In a feasible implementation manner, determining the second influence weight of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted includes:

[0014] Aggregate various user behaviors corresponding to each commodity category in the second historical behavior data at different sampling time periods respectively to obtain an aggregation result, where the aggregation result includes a three-dimensional matrix. The first dimension of the three-dimensional matrix is the user identifier of each user, the second dimension is various user behaviors corresponding to each commodity category, and the third dimension is the sampling time period;

[0015] Input the aggregation result into a preset Long Short-Term Memory (LSTM) model for training to determine the second influence weight.

[0016] In a feasible implementation manner, the user behavior sequence is formed by arranging the user behaviors corresponding to the current commodity category within the second time period in chronological order; the first time period includes the second time period, and the duration of the second time period is less than the duration of the first time period.

[0017] In a feasible implementation manner, according to the first influence weight and the second influence weight, determining the target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted includes:

[0018] Input the first influence weight and the second influence weight into a preset fully connected neural network layer to obtain the target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted.

[0019] In a feasible implementation manner, after determining the potential users corresponding to the commodity category to be predicted among the multiple users, it further includes:

[0020] Push the commodities of the commodity category to be predicted to the potential users online and collect the behavior data of the potential users;

[0021] Save the collected behavioral data of the potential users to the database and update the database;

[0022] Based on the updated database, re-determine the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted until the re-determined target influence weights are greater than or equal to a preset threshold.

[0023] In a second aspect, an embodiment of the present application provides a prediction device for potential users, and the device includes:

[0024] A first training module, configured to determine, according to first historical behavior data of multiple users on each commodity category within a first time period, first influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first historical behavior data;

[0025] A second training module, configured to determine, according to second historical behavior data of the multiple users on each commodity category within a second time period, second influence weights of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted according to the second historical behavior data;

[0026] A processing module, configured to determine target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first influence weights and the second influence weights;

[0027] A prediction module, configured to determine potential users corresponding to the commodity category to be predicted among the multiple users according to the target influence weights.

[0028] In a feasible implementation manner, the first training module is specifically configured to:

[0029] Aggregate various user behaviors corresponding to each commodity category in the first historical behavior data to obtain an aggregation result, where the aggregation result includes a two-dimensional matrix, the first dimension of the two-dimensional matrix is the user identifier of each user, and the second dimension is various user behaviors corresponding to each commodity category;

[0030] Input the aggregation result into a preset DNN model for training to determine the first influence weights.

[0031] In a feasible implementation manner, the second training module is specifically configured to:

[0032] Aggregate the various user behaviors corresponding to each commodity category in the second historical behavior data during different sampling time periods respectively to obtain an aggregation result, where the aggregation result includes a three-dimensional matrix, the first dimension of the three-dimensional matrix is the user identifier of each user, the second dimension is the various user behaviors corresponding to each commodity category, and the third dimension is the sampling time period;

[0033] Input the aggregation result into a preset LSTM model for training to determine the second influence weight.

[0034] In a feasible implementation manner, the processing module is specifically configured to:

[0035] Input the first influence weight and the second influence weight into a preset fully connected neural network layer to obtain the target influence weight of the various user behaviors corresponding to each commodity category on the commodity category to be predicted.

[0036] In a feasible implementation manner, it further includes a practical test module for:

[0037] After determining the potential users corresponding to the commodity category to be predicted among the multiple users, push the commodities of the commodity category to be predicted to the potential users online and collect the behavior data of the potential users;

[0038] Save the behavior data of the collected potential users to the database and update the database;

[0039] Based on the updated database, re-determine the target influence weight of the various user behaviors corresponding to each commodity category on the commodity category to be predicted until the re-determined target influence weight is greater than or equal to a preset threshold.

[0040] In a third aspect, an embodiment of the present application provides an electronic device, including: at least one processor and a memory;

[0041] The memory stores computer execution instructions;

[0042] The at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the prediction method of potential users provided in the first aspect.

[0043] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer execution instructions are stored, and when the processor executes the computer execution instructions, the prediction method of potential users provided in the first aspect is implemented.

[0044] The prediction method and device for potential users provided by the embodiments of the present application, based on the overall historical behavior of users, consider the user behavior and behavior sequence of users in a recent period of time, so as to obtain the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and then accurately predict the potential users corresponding to the commodity category to be predicted, improving the accuracy of the potential user prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the description of the embodiments of the present application or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 It is a schematic flowchart of a prediction method for potential users provided by an embodiment of the present application;

[0047] Figure 2 It is a schematic diagram of a deep learning neural network model provided in an embodiment of the present application;

[0048] Figure 3 It is a schematic flowchart of another prediction method for potential users provided by an embodiment of the present application;

[0049] Figure 4 It is a schematic diagram of program modules of a prediction device for potential users provided in an embodiment of the present application;

[0050] Figure 5 It is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application. In addition, although the disclosed content in the present application is introduced according to exemplary one or several examples, it should be understood that each aspect of these disclosed contents can also be independently constituted as a complete implementation manner.

[0052] It should be noted that the brief description of terms in this application is only for the convenience of understanding the following-described embodiments, rather than intending to limit the embodiments of this application. Unless otherwise specified, these terms should be understood in their ordinary and common meanings.

[0053] In this application, terms such as "first", "second", etc. in the specification, claims, and the above-mentioned drawings are used to distinguish similar or like objects or entities, and do not necessarily imply limiting a specific order or sequence, unless otherwise noted. It should be understood that such terms can be interchanged under appropriate circumstances, for example, it is possible to implement in an order other than those given in the illustration or description of the embodiments of this application.

[0054] In addition, the terms "comprising" and "having" and any variations thereof are intended to cover but not be exclusive of inclusion. For example, a product or device comprising a series of components does not necessarily have to be limited to those components clearly listed, but may include other components not clearly listed or inherent to these products or devices.

[0055] The term "module" used in this application refers to any known or later-developed hardware, software, firmware, artificial intelligence, fuzzy logic, or a combination of hardware and / or software code that can perform functions related to that element.

[0056] In the technical solution of this application, the processing of collection, storage, use, processing, transmission, provision, and disclosure of information such as user data complies with the provisions of relevant laws and regulations and does not violate public order and good customs.

[0057] In the field of e-commerce, at present, refined and precise marketing is required, that is, to find users with high purchase potential for targeted marketing. Then, how to accurately and effectively find users with high purchase potential has become a technical problem that urgently needs to be solved at present.

[0058] In the process of a user selecting and purchasing goods on an e-commerce platform, various operation behaviors such as browsing, collecting, adding to the shopping cart, and purchasing are usually accompanied. By deeply analyzing these operation behaviors, the shopping habits, preferences, or shopping needs of the user can be predicted, and then based on the prediction results, goods can be recommended to the user in a targeted manner, which can increase the sales volume of the goods and at the same time provide a better shopping experience for the user.

[0059] In traditional technologies, generally, based on the historical operation behaviors of users, traditional machine learning methods such as statistics and regression analysis are used to mine and predict potential users who may purchase a certain commodity.

[0060] However, in actual situations, the user behaviors closer to the operation time have a greater impact on the prediction results. Additionally, there is a certain correlation between user behaviors before and after, and the order of user behaviors will significantly affect the final prediction results. However, the current technology only macroscopically considers the overall historical behaviors of users and does not distinguish user behaviors based on time and operation order, resulting in inaccurate prediction results.

[0061] In the face of the above technical problems, in the embodiments of the present application, a method for predicting user purchase behaviors is provided. This method provides a new deep learning neural network model for predicting high-potential users. On the basis of considering the overall historical behaviors of users, it separately calculates the user behaviors and behavior order of users in the recent period using a dedicated model, so as to obtain the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and then accurately predicts the potential users corresponding to the commodity category to be predicted, improving the accuracy of the potential user prediction results. The following will be described in detail with specific embodiments.

[0062] Refer to Figure 1 , Figure 1 which is a schematic flowchart of a method for predicting potential users provided by the embodiments of the present application. As Figure 1 shown, the method for predicting potential users includes:

[0063] S101. Determine the first influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first historical behavior data of multiple users for each commodity category within the first time period.

[0064] In the embodiments of the present application, the operation behavior data of all or multiple users on the shopping platform can be collected in real time by a terminal device or a server. The operation behavior data includes, but is not limited to: operation behaviors such as browsing, favoriting, adding to the shopping cart, purchasing, sharing, and searching for a certain commodity.

[0065] In a feasible implementation manner, the collected operation behavior data of multiple users is saved in a preset database.

[0066] Optionally, the collected operation behavior data of multiple users can be stored in a Hadoop cluster.

[0067] Among them, Hadoop is a distributed system infrastructure that can perform high-speed operations and storage. Specifically, Hadoop implements a distributed file system (Distributed File System), and one of its components is HDFS (Hadoop Distributed File System). HDFS has the characteristics of high fault tolerance and is designed to be deployed on low-cost hardware; moreover, it provides high throughput to access application data and is suitable for applications with extremely large datasets.

[0068] In some embodiments, the first historical behavior data of multiple users for each commodity category within the first time period can be obtained from the Hadoop cluster by means of hive data reading, and then, based on the first historical behavior data, the first influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted can be determined.

[0069] Among them, the above-mentioned first time period can start from the moment when user behaviors are collected and end at the current moment. Alternatively, the above-mentioned first time period can also be determined according to a preset duration, such as the most recent one year, the most recent two years, etc., and there is no limitation in the embodiments of the present application.

[0070] To better understand the embodiments of the present application, assume that the commodity to be predicted is a mobile phone. Then, through the above-mentioned first historical behavior data, various user behaviors corresponding to each commodity category, such as adding clothes to the shopping cart, purchasing headphones, and favoriting mobile phones, etc., are determined for their influence weights on the user's purchase of mobile phones.

[0071] It can be understood that mobile phones belong to electronic products. Therefore, the behavior of a user favoriting a mobile phone has a relatively large influence weight on the user's purchase of a mobile phone. Secondly, the behavior of purchasing headphones also has an influence, while the behavior of adding clothes to the shopping cart has a relatively small influence weight on the user's purchase of a mobile phone.

[0072] S102. Determine the second influence weights of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted according to the second historical behavior data of multiple users for each commodity category within the second time period.

[0073] Among them, the above-mentioned user behavior sequence is formed by arranging the user behaviors corresponding to the current commodity category within the second time period in chronological order.

[0074] It can be understood that in actual situations, user behaviors closer to the operation time have a greater influence on the prediction result; in addition, there is a certain correlation between user behaviors, and the order of user behaviors will significantly affect the final prediction result.

[0075] For example, when a user has the behaviors of browsing, favoriting, and adding to the shopping cart for a certain product in sequence, it is highly likely that the user will purchase the product. Therefore, when a user has the behaviors of browsing, favoriting, and unfavoriting a certain product in sequence, it is highly unlikely that the user will purchase the product.

[0076] For another example, the operation of favoriting headphones by the user three days ago has a greater impact weight on the user's purchase of headphones than the operation of favoriting headphones in the previous thirty days.

[0077] Among them, the above-mentioned first time period includes the above-mentioned second time period, and the duration of the second time period is less than the duration of the first time period.

[0078] Exemplarily, when the first time period is the most recent year, the second time period can be the most recent month.

[0079] To better understand the embodiments of the present application, assume that the product to be predicted is a mobile phone. Then, through the above-mentioned second historical behavior data, various user behavior sequences corresponding to each product category are determined, such as user behavior sequences of browsing headphones, favoriting headphones, etc. Then, the influence weights of these user behavior sequences on the purchase of mobile phones are determined.

[0080] S103. According to the first influence weight and the second influence weight, determine the target influence weight of various user behaviors corresponding to each product category on the product category to be predicted, and according to the target influence weight, determine the potential users corresponding to the product category to be predicted among multiple users.

[0081] In some embodiments, a fully connected neural network layer can be used to aggregate the above-mentioned first influence weight and the second influence weight to obtain the target influence weight of various user behaviors corresponding to each product category on the product category to be predicted.

[0082] In some embodiments, the above-mentioned target influence weight includes the influence weight of various user behaviors corresponding to each product category on the product category to be predicted, and the influence weight of various user behavior sequences corresponding to each product category on the product category to be predicted.

[0083] In some embodiments, according to the above-mentioned first historical behavior data and the target influence weight, a Softmax classifier can be used to classify the above-mentioned multiple users into two categories, one category is the potential users corresponding to the product category to be predicted, and the other category is the non-potential users corresponding to the product category to be predicted.

[0084] The prediction method for potential users provided by the embodiments of the present application, based on the overall historical behavior of users, considers the user behavior and behavior sequence of users in the recent period, so as to obtain the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and then accurately predict the potential users corresponding to the commodity category to be predicted, improving the accuracy of the potential user prediction result.

[0085] Based on the content described in the above embodiments, in a feasible implementation manner, after obtaining the first historical behavior data of multiple users for each commodity category in the first recent time period in the database, aggregate various user behaviors corresponding to each commodity category in the first historical behavior data to obtain an aggregation result, and the aggregation result includes a two-dimensional matrix, where the first dimension of the two-dimensional matrix is the user identifier of each user, and the second dimension is various user behaviors corresponding to each commodity category.

[0086] For the browsing, attention, adding to the shopping cart, purchasing and other operation behaviors of users for all commodities in history, aggregate them according to commodity category + operation type, and all users form a two-dimensional matrix of user identifier * feature, where a feature identifier represents a user behavior corresponding to a commodity category, for example, "collecting clothes" can be used as a feature.

[0087] In some embodiments, the above two-dimensional matrix can be input into a 4-layer DNN model and trained according to the first historical behavior data to obtain the first influence weights of each category feature on the commodity category to be predicted.

[0088] In a feasible implementation manner, after obtaining the second historical behavior data of multiple users for each commodity category in the second recent time period in the database, the various user behaviors corresponding to each commodity category in different sampling time periods in the second historical behavior data can be aggregated respectively to obtain an aggregation result, and the aggregation result includes a three-dimensional matrix, where the first dimension of the three-dimensional matrix is the user identifier of each user, the second dimension is various user behaviors corresponding to each commodity category, and the third dimension is the sampling time period.

[0089] It can be understood that the user behavior closer to the present has a greater impact on the prediction result. Therefore, the recent historical behavior of users can form a user behavior sequence according to the time distance. Using the LSTM model to process this user behavior sequence, the second influence weights of user behaviors at different time distances on the commodity category to be predicted can be obtained.

[0090] Exemplarily, the user behaviors in a recent period (such as 30 days) can be aggregated and statistically analyzed by day granularity and characterized by behavior types to obtain a three-dimensional matrix of user *features* and dates, and then input into the LSTM model for training to determine the second influence weight of the user behavior sequence in the recent period on the product category to be predicted.

[0091] In some embodiments, the first influence weight and the second influence weight are aggregated together and aggregated into a unified target influence weight through a fully connected neural network layer, and the target influence weight includes the influence weights of all features on the product category to be predicted.

[0092] In some embodiments, according to the above first historical behavior data and the target influence weight, using a Softmax classifier, the above multiple users can be divided into two categories, one category is potential users corresponding to the product category to be predicted, and the other category is non-potential users corresponding to the product category to be predicted.

[0093] For a better understanding of the embodiments of the present application, refer to Figure 2 , Figure 2 which is a schematic diagram of a deep learning neural network model provided in the embodiments of the present application.

[0094] In some embodiments of the present application, the DNN model includes an input layer, a hidden layer 1, a hidden layer 2, and an output layer. f1, f2, ……, fn are the input data of the DNN model, and r1, r2 are the output data of the DNN model.

[0095] The first step of the LSTM model is to determine which information to discard from the cell state, and this decision is made by a sigmoid layer called the "forget gate layer", which looks at Ht and Dt and outputs a number between 0 and 1 for each number in the cell state Ct. 1 represents "keep this completely", while 0 represents "get rid of this completely".

[0096] Among them, when the second recent period is 30 days, Dt-30 can be understood as the 30th day recently, and Ht-30 and Dt-30 can be understood as the user behavior data on the 30th day recently; Dt-29 can be understood as the 29th day recently, and Ht-29 and Dt-29 can be understood as the user behavior data on the 29th day recently; ……, Dt-1 can be understood as the 1st day recently, and Ht-1 and Dt-1 can be understood as the user behavior data on the 1st day recently.

[0097] The prediction method for potential users provided by the embodiments of the present application combines DNN and LSTM neural network models. Based on the overall historical behavior of users, it considers the user behavior and behavior sequence of users in a recent period of time, so as to obtain the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and then accurately predicts the potential users corresponding to the commodity category to be predicted, improving the accuracy of the potential user prediction results.

[0098] Based on the content described in the above embodiments, refer to Figure 3 , Figure 3 which is a schematic flowchart of another prediction method for potential users provided by the embodiments of the present application. As Figure 3 shown, the prediction method for potential users includes:

[0099] Obtain the first historical behavior data of multiple users for each commodity category within the first recent time period in the database, and the second historical behavior data of the multiple users for each commodity category within the second recent time period.

[0100] Through model training with DNN and LSTM neural network models, determine the first influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and the second influence weights of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted. According to the first influence weights and the second influence weights, determine the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted.

[0101] Based on the above first historical behavior data and target influence weights, determine the potential users corresponding to the commodity category to be predicted among multiple users.

[0102] Conduct a practical test, that is, push the commodities of the commodity category to be predicted to potential users online and collect the behavior data of potential users.

[0103] Save the collected behavior data of potential users to the database and update the database.

[0104] Based on the updated database, re-determine the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted until the re-determined target influence weights are greater than or equal to a preset threshold.

[0105] The prediction method for potential users provided by the embodiments of the present application, based on the overall historical behavior of users, considers the user behavior and behavior sequence of users in a recent period of time, so as to obtain the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and then accurately predicts the potential users corresponding to the commodity category to be predicted, improving the accuracy of the potential user prediction results.

[0106] Based on the content described in the above embodiments, an apparatus for predicting potential users is further provided in an embodiment of the present application. Referring to Figure 4 , Figure 4 which is a schematic diagram of program modules of an apparatus for predicting potential users provided in an embodiment of the present application. The apparatus for predicting potential users includes:

[0107] A first training module 401, configured to determine a first influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to first historical behavior data of multiple users for each commodity category within a first time period.

[0108] A second training module 402, configured to determine a second influence weight of various user behavior sequences corresponding to each commodity category on the commodity category to be predicted according to second historical behavior data of multiple users for each commodity category within a second time period.

[0109] Wherein, the user behavior sequence is formed by arranging user behaviors corresponding to the current commodity category within the second time period in chronological order; the first time period includes the second time period, and the duration of the second time period is less than the duration of the first time period.

[0110] A processing module 403, configured to determine a target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted according to the first influence weight and the second influence weight.

[0111] A prediction module 404, configured to determine potential users corresponding to the commodity category to be predicted among multiple users according to the target influence weight.

[0112] The apparatus for predicting potential users provided in the embodiment of the present application, based on the overall historical behavior of users, considers the user behavior and behavior sequence of users in the recent period of time, so as to obtain the target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted, and further accurately predict the potential users corresponding to the commodity category to be predicted, improving the accuracy of the potential user prediction result.

[0113] In a feasible implementation manner, the first training module 401 is specifically configured to:

[0114] Aggregate various user behaviors corresponding to each commodity category in the first historical behavior data to obtain an aggregation result, where the aggregation result includes a two-dimensional matrix, the first dimension of the two-dimensional matrix is the user identifier of each user, and the second dimension is various user behaviors corresponding to each commodity category.

[0115] Input the above aggregation result into a preset DNN model for training to determine the first influence weight.

[0116] In a feasible implementation manner, the second training module 402 is specifically configured to:

[0117] Aggregate various user behaviors corresponding to each commodity category in the second historical behavior data within different sampling time periods respectively to obtain an aggregation result, where the aggregation result includes a three-dimensional matrix, the first dimension of the three-dimensional matrix is the user identifier of each user, the second dimension is various user behaviors corresponding to each commodity category, and the third dimension is the sampling time period.

[0118] Input the above aggregation result into a preset LSTM model for training to determine the second influence weight.

[0119] In a feasible implementation manner, the processing module 403 is specifically configured to:

[0120] Input the first influence weight and the second influence weight into a preset fully connected neural network layer to obtain the target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted.

[0121] In a feasible implementation manner, it further includes a practical test module, which is used for:

[0122] After determining potential users corresponding to the commodity category to be predicted among multiple users, push the commodities of the commodity category to be predicted to the potential users online and collect the behavior data of the potential users; save the collected behavior data of the potential users to the database and update the database; based on the updated database, re-determine the target influence weight of various user behaviors corresponding to each commodity category on the commodity category to be predicted until the re-determined target influence weight is greater than or equal to a preset threshold.

[0123] It should be noted that the specific contents executed by the first training module 401, the second training module 402, the processing module 403, and the prediction module 404 in the embodiments of the present application can refer to Figures 1 to 3 the relevant contents in the embodiments shown, which will not be elaborated here.

[0124] Furthermore, based on the content described in the above embodiments, an user device is further provided in the embodiments of the present application. The user device includes at least one processor and a memory; wherein, the memory stores computer execution instructions; the above at least one processor executes the computer execution instructions stored in the memory to implement each step of the prediction method of potential users as described in the above embodiments, and will not be elaborated here in this embodiment.

[0125] For a better understanding of the embodiments of the present application, refer to Figure 5 , Figure 5 which is a schematic hardware structure diagram of an electronic device provided for the embodiments of the present application.

[0126] AsFigure 5 As shown in Figure 5 , the electronic device 50 of this embodiment includes: a processor 501 and a memory 502; where:

[0127] The memory 502 is used to store computer-executable instructions;

[0128] The processor 501 is used to execute the computer-executable instructions stored in the memory to implement the various steps of the method for predicting potential users described in the above embodiments, which will not be elaborated here in this embodiment.

[0129] Optionally, the memory 502 can be either independent or integrated with the processor 501.

[0130] When the memory 502 is independently provided, the device further includes a bus 503 for connecting the memory 502 and the processor 501.

[0131] Furthermore, based on the content described in the above embodiments, an embodiment of the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the processor executes the computer-executable instructions, the various steps of the method for predicting potential users described in the above embodiments are implemented, which will not be elaborated here in this embodiment.

[0132] In several embodiments provided by the present application, it should be understood that the disclosed device and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces, and the indirect coupling or communication connection of the device or module can be in an electrical, mechanical or other form.

[0133] The modules described as separate components may or may not be physically separated. The components displayed as modules may or may not be physical units, that is, they can be located in one place, or they can be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0134] In addition, in each embodiment of the present application, the various functional modules can be integrated in a processing unit, or each module can exist physically alone, or two or more modules can be integrated in one unit. The above-mentioned unit integrated with modules can be implemented in the form of hardware, or in the form of a hardware plus software functional unit.

[0135] The integrated modules implemented in the form of software functional modules can be stored in a computer-readable storage medium. The above-mentioned software functional modules stored in a storage medium include several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to execute some steps of the methods described in various embodiments of the present application.

[0136] It should be understood that the above-mentioned processor may be a central processing unit (English: Central Processing Unit, abbreviated as: CPU), and may also be other general-purpose processors, digital signal processors (English: Digital Signal Processor, abbreviated as: DSP), application-specific integrated circuits (English: Application Specific Integrated Circuit, abbreviated as: ASIC), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0137] The memory may include high-speed RAM memory, and may also include non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk, or an optical disc, etc.

[0138] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the buses in the drawings of the present application are not limited to only one bus or one type of bus.

[0139] The above-mentioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disc. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0140] An exemplary storage medium is coupled to a processor, enabling the processor to read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an Application Specific Integrated Circuits (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a master device.

[0141] Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disks, or optical discs that can store program codes.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting potential users, characterized in that, The method includes: Determining a first influence weight of various user behaviors corresponding to each commodity category on a to-be-predicted commodity category according to first historical behavior data of multiple users for each commodity category within a first time period; Determining a second influence weight of various user behavior sequences corresponding to each commodity category on the to-be-predicted commodity category according to second historical behavior data of the multiple users for each commodity category within a second time period; Determining a target influence weight of various user behaviors corresponding to each commodity category on the to-be-predicted commodity category according to the first influence weight and the second influence weight, and determining potential users corresponding to the to-be-predicted commodity category among the multiple users according to the target influence weight; The determining the first influence weight of various user behaviors corresponding to each commodity category on the to-be-predicted commodity category includes: Aggregating various user behaviors corresponding to each commodity category in the first historical behavior data to obtain an aggregation result, where the aggregation result includes a two-dimensional matrix, the first dimension of the two-dimensional matrix is the user identifier of each user, and the second dimension is various user behaviors corresponding to each commodity category; Inputting the aggregation result into a preset deep neural network DNN model for training to determine the first influence weight; The determining the second influence weight of various user behavior sequences corresponding to each commodity category on the to-be-predicted commodity category includes: Respectively aggregating various user behaviors corresponding to each commodity category in different sampling time periods in the second historical behavior data to obtain an aggregation result, where the aggregation result includes a three-dimensional matrix, the first dimension of the three-dimensional matrix is the user identifier of each user, the second dimension is various user behaviors corresponding to each commodity category, and the third dimension is the sampling time period; Inputting the aggregation result into a preset long short-term memory LSTM model for training to determine the second influence weight; The determining the target influence weight of various user behaviors corresponding to each commodity category on the to-be-predicted commodity category according to the first influence weight and the second influence weight includes: Inputting the first influence weight and the second influence weight into a preset fully connected neural network layer to obtain the target influence weight of various user behaviors corresponding to each commodity category on the to-be-predicted commodity category.

2. The method according to claim 1, characterized in that, The user behavior sequence is formed by arranging user behaviors corresponding to the current commodity category in chronological order within the second time period; the first time period includes the second time period, and the duration of the second time period is less than the duration of the first time period.

3. The method according to claim 1 or 2, characterized in that, After determining the potential users corresponding to the to-be-predicted commodity category among the multiple users, it further includes: Online pushing the commodity of the to-be-predicted commodity category to the potential users and collecting the behavior data of the potential users; Saving the collected behavior data of the potential users to a database and updating the database; Based on the updated database, re-determine the target influence weights of various user behaviors corresponding to each product category on the product category to be predicted until the re-determined target influence weights are greater than or equal to a preset threshold.

4. A device for predicting potential users, characterized in that, The device includes: A first training module, configured to determine the first influence weights of various user behaviors corresponding to each product category on the product category to be predicted according to the first historical behavior data of multiple users for each product category within a first time period; A second training module, configured to determine the second influence weights of various user behavior sequences corresponding to each product category on the product category to be predicted according to the second historical behavior data of the multiple users for each product category within a second time period; A processing module, configured to determine the target influence weights of various user behaviors corresponding to each product category on the product category to be predicted according to the first influence weights and the second influence weights; A prediction module, configured to determine potential users corresponding to the product category to be predicted among the multiple users according to the target influence weights; The first training module is specifically configured to Aggregate various user behaviors corresponding to each product category in the first historical behavior data to obtain an aggregation result, where the aggregation result includes a two-dimensional matrix, the first dimension of the two-dimensional matrix is the user identifier of each user, and the second dimension is various user behaviors corresponding to each product category; Input the aggregation result into a preset deep neural network DNN model for training to determine the first influence weights; The second training module is specifically configured to Respectively aggregate various user behaviors corresponding to each product category in different sampling time periods in the second historical behavior data to obtain an aggregation result, where the aggregation result includes a three-dimensional matrix, the first dimension of the three-dimensional matrix is the user identifier of each user, the second dimension is various user behaviors corresponding to each product category, and the third dimension is the sampling time period; Input the aggregation result into a preset long short-term memory LSTM model for training to determine the second influence weights; The processing module is specifically configured to input the first influence weights and the second influence weights into a preset fully connected neural network layer to obtain the target influence weights of various user behaviors corresponding to each product category on the product category to be predicted.

5. The device according to claim 4, characterized in that, The user behavior sequence is formed by arranging the user behaviors corresponding to the current product category within the second time period in chronological order; The first time period includes the second time period, and the duration of the second time period is less than the duration of the first time period.

6. The device according to claim 4 or 5, characterized in that, It further includes a practical test module, configured to: After determining potential users corresponding to the product category to be predicted among the multiple users, push the products of the product category to be predicted to the potential users online and collect the behavior data of the potential users; Save the behavior data of the collected potential users to the database and update the database; Based on the updated database, re-determine the target influence weights of various user behaviors corresponding to each commodity category on the commodity category to be predicted until the re-determined target influence weights are greater than or equal to a preset threshold.

7. An electronic device, characterized in that, Including: At least one processor and a memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the prediction method for potential users according to any one of claims 1 to 3.

8. A computer-readable storage medium, characterized in that, Computer-executable instructions are stored in the computer-readable storage medium, and when the processor executes the computer-executable instructions, the prediction method for potential users according to any one of claims 1 to 3 is implemented.

Images