WO2019179030A1 - Product purchasing prediction method, server and storage medium - Google Patents

Abstract

A product purchasing prediction method, a server and a storage medium, the method comprising: receiving an analysis request which carries identity label information and a target product type (S10); extracting purchasing data of a target user and target product purchasing data of other users (S20); generating a corresponding purchasing time sequence according to purchasing time points (S30); inputting, according to mapping between the product type and a first analysis model, the purchasing time sequence of the target user into the first analysis model so as to generate a first purchasing prediction value (S40); performing mean value processing on the purchasing time sequence for the target product of a preset amount of users, and generating an average purchasing time sequence of the preset amount of users for the target product (S50); after performing mean value processing on the purchasing time sequences of the other users, inputting same into a first analysis model to generate a second purchasing prediction value (S60); and according to mapping between product types and second analysis models, inputting the first purchasing prediction value and the second purchasing prediction value for the target product as well as first purchasing prediction values of product types other than the target product type into a corresponding second analysis model so as to generate a final purchasing prediction value for the target product (S70), and thus improving the accuracy of user purchasing intention prediction.

Description

Product purchase forecasting method, server and storage medium

This application claims priority to Chinese Patent Application No. 201101226460.4, entitled "Product Purchase Prediction Method, Server and Storage Media", filed on March 19, 2018, the entire contents of which are incorporated by reference. The way it is combined with this application.

Technical field

The present application relates to the field of product purchase prediction, and in particular, to a product purchase prediction method, a server, and a computer readable storage medium.

Background technique

In the financial business, it is necessary to predict in advance the user's propensity to purchase financial products such as microinsurance, stocks, funds, etc., which is conducive to the marketing and service of financial product marketing parties in advance.

The existing propensity analysis and prediction method usually uses cross-sectional data to analyze the user's purchasing tendency. This method can only predict the user's purchasing tendency in one time, and cannot predict the user's continuous purchasing tendency and the user's point preference at the time of purchase. .

Summary of the invention

In view of the above, the present application provides a product purchase prediction method, a server, and a computer readable storage medium, the main purpose of which is to predict a user's continuous purchase tendency and improve the accuracy of financial product purchase prediction.

To achieve the above objective, the present application provides a product purchase prediction method, the method comprising:

Receiving step: receiving an analysis request with the target user identity information and the target product type;

The extracting step: extracting, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type;

Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.

In addition, the application further provides a server, comprising: a memory, a processor and a display, wherein the memory stores a product purchase prediction program, and the product purchase prediction program is executed by the processor, and the following steps can be implemented:

Receiving step: receiving an analysis request with the target user identity information and the target product type;

The extracting step: extracting, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type;

Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.

In addition, in order to achieve the above object, the present application further provides a computer readable storage medium including a product purchase prediction program, where the product purchase prediction program is executed by a processor, as described above Any step in the product purchase forecasting method.

The product purchase prediction method, the server and the computer readable storage medium proposed by the present application respectively extract various product purchases of the target user from a plurality of service servers by receiving an analysis request with the target user identity information and the target product category. Data and a preset number of other users' target product purchase data, and extracting a purchase time point to generate a purchase time series, and then inputting the target user's purchase time series into the corresponding first analysis model to generate a first purchase predicted value, The other user's purchase time series is input to the corresponding first analysis model, the second purchase predicted value is generated, and finally the first purchase predicted value and the second purchased predicted value are input into the second analysis model to generate a final purchase predicted value of the target product. This application enables the present application to predict the continuous purchase tendency of the target user and improve the prediction accuracy.

DRAWINGS

1 is a schematic diagram of a preferred embodiment of a server of the present application;

2 is a block diagram showing a preferred embodiment of the product purchase prediction program of FIG. 1;

3 is a flow chart of a first embodiment of a method for predicting product purchase according to the present application;

4 is a flow chart of a second embodiment of a method for predicting product purchase according to the present application;

Figure 5 is a flow chart of the training of the first analysis model of the present application;

6 is a flowchart of training of a second analysis model of the present application;

Figure 7 is a schematic diagram of the purchase data of the product of the present application.

The implementation, functional features and advantages of the present application will be further described with reference to the accompanying drawings.

detailed description

It is understood that the specific embodiments described herein are merely illustrative of the application and are not intended to be limiting.

As shown in FIG. 1, it is a schematic diagram of a preferred embodiment of the server 1 of the present application.

In this embodiment, the server 1 may be a server, a smart phone, a tablet computer, a personal computer, a portable computer, and other electronic devices having computing functions.

The server 1 includes a memory 11, a processor 12, a display 13, a network interface 14, and a communication bus 15. The network interface 14 can optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). Communication bus 15 is used to implement connection communication between these components.

The memory 11 includes at least one type of readable storage medium. The at least one type of readable storage medium may be a non-volatile storage medium such as a flash memory, a hard disk, a multimedia card, a card type memory, or the like. In some embodiments, the memory 11 may be an internal storage unit of the server 1, such as a hard disk of the server 1. In other embodiments, the memory 11 may also be an external storage unit of the server 1, such as a plug-in hard disk equipped on the server 1, a smart memory card (SMC), and a secure digital ( Secure Digital, SD) cards, flash cards, etc.

In this embodiment, the memory 11 can be used not only for storing application software installed on the server 1 and various types of data, such as a product purchase prediction program 10, an analysis request with target user identification information and a target product type. Wait.

The processor 12, in some embodiments, may be a Central Processing Unit (CPU), microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing product purchase predictions. The computer program code of the program 10, the training of the first analysis model and the second analysis model, and the like.

Display 13 can be referred to as a display screen or display unit. In some embodiments, the display 13 can be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an Organic Light-Emitting Diode (OLED) touch sensor. The display 13 is used to display information processed in the server 1 and a work interface for displaying visualizations, such as displaying a final purchase predicted value of the target product at the next purchase.

Figure 1 shows only server 1 with components 11-15 and product purchase prediction program 10, but it should be understood that not all illustrated components may be implemented, and more or fewer components may be implemented instead.

Optionally, the server 1 may further include a user interface, and the user interface may include an input unit such as a keyboard, a voice output device such as an audio, a headset, etc., optionally, the user interface may further include a standard wired interface and a wireless interface.

The server 1 may also include radio frequency (RF) circuits, sensors, audio circuits, and the like, and details are not described herein.

2 is a block diagram of a preferred embodiment of the product purchase prediction program 10 of FIG. A module as referred to in this application refers to a series of computer program instructions that are capable of performing a particular function.

In this embodiment, the product purchase prediction program 10 includes: a receiving module 110, an extracting module 120, a generating module 130, a first predicting module 140, an average processing module 150, a second prediction module 160, and a final prediction module 170. The functions or operational steps implemented by 110-170 are similar to the above, and are not described in detail herein, by way of example, for example:

The receiving module 110 is configured to receive an analysis request with target user identity information and a target product type, or receive purchase data of various product types corresponding to all users from a plurality of predetermined service servers in real time. The target user identity information is information indicating the identity of the target user, such as a user name and an identity card number. The product types include: stocks, funds, wealth management, insurance and other products.

The extracting module 120 is configured to separately extract, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type. The service server includes: a bank server, a stock server, an insurance server, and the like. The purchase data refers to the user identification information and the purchase time point of the corresponding product.

The generating module 130 generates a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents an interval of consecutive purchase behaviors of the same number of days, and if X is 0, it represents a non-continuous purchase. Behavior, Y represents the number of consecutive purchases of the same number of days.

The first prediction module 140 is configured to determine, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and input a purchase time sequence corresponding to each product separately. In the corresponding first analysis model, a first purchase prediction value corresponding to each product is generated. Each product type corresponds to a first analysis model, which is pre-built and trained. The first analysis model is a Long Short-Term Memory (LSTM) model, and the training process for the first analysis model corresponding to one product is as follows:

S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.

The averaging processing module 150 is configured to perform an averaging process on the purchase time series of the target product of the preset number of users, and generate an average purchase time sequence corresponding to the target product of the preset number of users. Wherein, the steps of the mean processing are as follows:

The average number of occurrences of consecutive purchase behaviors in the purchase time series of the target product purchased by the preset number of users is averaged, and the average number of consecutive purchase behaviors of all users for various preset time intervals of the target product is obtained, and the target product is obtained. The various preset time intervals and the corresponding average number of occurrences of the consecutive purchase behaviors constitute an average purchase time series corresponding to the target product.

The second prediction module 160 is configured to determine, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and input the average purchase time series to In the corresponding first analysis model, a second purchase prediction value of the target product is generated.

The final prediction module 170 is configured to determine, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and save the first purchase predicted value of the target product. A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the target product type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated. Wherein, the second analysis model is pre-built and trained. The first analysis model is a Granger model (Granger model), and the training process for a second analysis model corresponding to one product is as follows:

S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the second analysis model of the training by using each sample data in the verification set after the training is completed;

S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.

As shown in FIG. 3, it is a flowchart of the first embodiment of the product purchase prediction method of the present application.

In this embodiment, when the processor 12 executes the computer program of the product purchase prediction program 10 stored in the memory 11, the method for implementing the product purchase prediction includes: Step S10 - Step S70:

In step S10, the receiving module 110 receives an analysis request with the target user identity information and the target product type. The target user identity information is information indicating the identity of the target user, such as a user name and an identity card number. The product types include: stocks, funds, wealth management, insurance and other products. The target product type may refer to one or more product types, but it should be understood that the method steps described below predict the user's purchasing propensity only for one of the types of products purchased by the target user.

Step S20: The extracting module 120 extracts, according to the received request, the purchase data of the various product types corresponding to the target user identity information from the plurality of predetermined service servers, and the preset number of users related to the target product type. Purchase data. The service server includes: a bank server, a stock server, an insurance server, and the like. The purchase data refers to the user identification information and the purchase time point of the corresponding product. However, it should be understood that a user not only purchases one product, but the purchase data of the various product types refers to purchase data of all product types purchased by the target user. As shown in FIG. 7, it is a schematic diagram of the purchase data of the product of the present application. Assume that user 1 is the target user, and the fund product that user 1 purchased at the bank server is the target product, then the purchase time point 12 is the purchase data of the target product type. The bank server records the identity information of the user 1 and various time points of the fund purchase. Different types of products can be sold by the same business server. For example, user 1 can purchase fund products on a bank server or purchase wealth management products on a bank server. The preset number of users may be all users who purchase products, or may be a preset number of randomly selected users.

In step S30, the generating module 130 generates a purchase time series (X, Y) corresponding to each product according to the purchase time points corresponding to the various products in the extracted various product purchase data. For example, the purchase time of the target product's fund product purchase data is: 2017.1.1, 2017.1.5, 2017.1.9, 2017.2.8, 2017.2.15, 2017.2.18, 2017.2.21, 2017.2.24, then Generating the purchase time series (4, 3), (0, 1), (3, 4) of the user's fund product, wherein the number 4 in (4, 3) represents the interval between consecutive purchases of 4 days apart The number 3 represents the number of purchases of consecutive purchases separated by the same number of days, the number 0 in (0, 1) represents the non-continuous purchase behavior, and the number 1 represents the number of purchases of the non-continuous purchase behavior.

Step S40: Determine a first analysis model corresponding to each product purchased by the target user according to a mapping relationship between the product type and the pre-trained first analysis model, and the first prediction module 140 inputs the purchase time series corresponding to each product separately. In the corresponding first analysis model, a first purchase prediction value corresponding to each product is generated. Each product type corresponds to a first analysis model, which is pre-built and trained. The first analysis model is an LSTM model, as shown in FIG. 5, which is a flowchart of the first analysis model training of the present application. The training process for the first analysis model corresponding to a product is as follows:

S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence. For example, 100,000 users are randomly selected from the bank server, the fund product purchase data of each user is extracted, and the purchase time series of the fund products of each user is generated according to the corresponding purchase time point.

S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio. For example, a purchase time series of a fund product of 80% of users is used as a training set, and a purchase time series of fund products of the remaining 20% of users is used as a verification set.

S13. Train the first analysis model by using a purchase time series in the training set, and verify the accuracy of the first analysis model by using a purchase time series in the verification set after the training. For example, the training time series of the fund products of the training set of 80,000 users is input into the model training, the first analysis model is generated, and the purchase time series of the fund products of the 20,000 users is input to the first analysis generated. Accuracy verification is performed in the model.

S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12. For example, if the verification accuracy is greater than 96%, the training is completed. If the accuracy is less than 96%, the purchase time sequence of 20,000 users is increased, and then step S12 is performed.

In step S50, the average processing module 150 performs an average processing on the purchase time series of the target product of the preset number of users, and generates an average purchase time sequence corresponding to the target product of the preset number of users. For example, the purchase time series of the fund products of 10,000 users are averaged to generate an average purchase time series corresponding to the fund products. Wherein, the steps of the mean processing are as follows:

The average number of occurrences of consecutive purchase behaviors in the purchase time series of the target product purchased by the preset number of users is averaged, and the average number of consecutive purchase behaviors of all users for various preset time intervals of the target product is obtained. The various preset time intervals of the product and the average number of occurrences of the corresponding consecutive purchase behaviors constitute an average purchase time sequence corresponding to the target product. For example, the purchase time series of the fund products of 10,000 users are arranged in a list, as shown in Table 1, wherein interval 0 represents a non-continuous purchase behavior, interval 1 represents a purchase time interval of 1 day, and interval N represents a purchase time. The interval is N days.

Table 1

User ID	product type	Interval 0	Interval 1	Interval 2	......	Interval N
ID00001	fund	7	10	4	......	3
ID00002	fund	4	5	6	......	2
......	......	......	......	......	......	......
ID10000	fund	3	5	4	......	0

The averaging processing module 150 averages the data in the columns of the various preset time intervals of the product, rounds the average value and then takes the integer, and obtains the average consecutive purchase times of all the preset time intervals of the user for the product. The various preset time intervals of the product and their corresponding average consecutive purchases constitute an average purchase time series for the corresponding product. For example, the average number of purchases at interval 0 is averaged: (7+4+...+3)/10000, assuming that the average number of purchases is 5, an average purchase time series for the generated fund product is (0,5). ).

Step S60: The second prediction module 160 inputs the average purchase time sequence corresponding to the target product into the corresponding first analysis model according to the mapping relationship between the product type and the pre-trained first analysis model, and generates a corresponding Second, purchase the predicted value. For example, the average purchase time series corresponding to the fund products of 10,000 users is input into the first analysis model corresponding to the fund products, and the second purchase predicted value of the fund products is generated.

Step S70: Determine, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the final prediction module 170 divides the first purchase prediction value corresponding to the target product. A second purchase model corresponding to the first purchase predicted value corresponding to the other type of products other than the target product type and the second purchase predicted value corresponding to the target product is input, and a final purchase predicted value corresponding to the target product is generated. For example, inputting a first purchase predicted value of the target user's fund product, a first purchase predicted value of the other type of product of the target user other than the fund product, and a second purchase predicted value of the fund product into the second analysis model of the fund product, Get the final purchase forecast for the fund product. Wherein, the second analysis model is pre-built and trained. The first analysis model is a Granger model, as shown in FIG. 6 , which is a flowchart of the second analysis model training of the present application, and the training process for the second analysis model corresponding to one product is as follows:

S21. Extracting purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generating corresponding products for each user according to corresponding purchase time points in the purchase data of the various products. Purchase time series. For example, various product purchase data corresponding to 100,000 users are extracted from a plurality of business servers such as a bank server, a stock server, and an insurance server, and a purchase time series corresponding to each product is generated based on the purchase time point in the purchase data.

S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model inputs the purchase time series corresponding to the various products purchased by the user into the corresponding first analysis model, and generates first purchase prediction values corresponding to the various products. The user selection one by one means that each user is selected as the target user one by one, and the first purchase prediction value of the user is calculated, until all the preset number of users are selected, and the first purchase prediction values of all the users are calculated. The purchase time series of each product of the preset number of users are averaged, respectively, and an average purchase time series corresponding to each product is generated, and an average purchase time series corresponding to each product is input to the corresponding first analysis. In the model, a second purchase predicted value corresponding to each product is generated. For example, the purchase time series of each product of 100,000 users is averaged, and an average purchase time series corresponding to each product is generated, and input into the first analysis model to obtain a second purchase prediction value of each product. The selected user is the first purchase predicted value for each product, the first purchase predicted value corresponding to other types of products other than the product, and the second purchase predicted value corresponding to the product of the product type as the user. Sample data for this product.

S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio. For example, 80% of the sample data is randomly used as the training set, and the remaining 20% of the sample data is used as the verification set.

S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the trained second analysis model by using each sample data in the verification set after the training is completed. For example, the sample data of 80,000 users in the training set is input into the model for training, a second analysis model is generated, and sample data of 20,000 users in the verification set is input into the generated first analysis model for accuracy verification.

S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of sample data is increased by increasing the number of users, and then step S22 is performed. For example, if the verification accuracy is greater than 98%, the training is completed. If the accuracy is less than 98%, the purchase time sequence of 20,000 users is increased, and then step S22 is performed.

The product purchase prediction method according to the above embodiment, according to the time series of the user purchasing the product, using the first analysis model to respectively calculate the first purchase prediction value of the target user and the second purchase prediction value of the target product, and finally the target user. The first purchase predicted value of the target product, the first purchase predicted value of the target product other than the target product, and the second purchase predicted value of the target product are input into the second analysis model to obtain the final target product of the target user. Buying predicted values, so as to accurately predict the user's continuous purchase tendency, and do a good job in marketing services in advance.

As shown in FIG. 4, it is a flowchart of the second embodiment of the product purchase prediction method of the present application.

In this embodiment, the product purchase prediction method includes: step S100 - step S700. The steps S300-S700 are substantially the same as those in the first embodiment, and details are not described herein again.

In step S100, the receiving module 110 receives various product purchase data corresponding to all users from a plurality of predetermined service servers in real time. For example, whenever a user makes a product purchase at a bank server, a stock server, or an insurance server, the receiving module 110 will automatically receive data from the bank server, stock server, or insurance server that the user purchased a product.

Step S200, after receiving the analysis request with the target user identity information and the target product type, the extraction module 120 extracts various product purchase data corresponding to the target user identity information and a preset number of users related to the target product purchase. data. For example, after receiving the analysis request of the target user identity information and the wealth management product, the extraction module 120 extracts various product purchase data corresponding to the user identity identification information and the wealth management product purchase data of 10,000 users.

The product purchase prediction method proposed in this embodiment receives the product purchase data of the user in real time, and when receiving the analysis request, extracts product purchase data of the target user and other users according to the request information, and generates a purchase time sequence, according to the purchase time series. Using the first analysis model, respectively calculating the first purchase predicted value of the target user and the second purchase predicted value of the target product, and finally, the first purchase predicted value of the target product of the target user, and the target user other products than the target product. The first purchase predicted value and the second purchase predicted value of the target product are input into the second analysis model to obtain a final purchase predicted value of the target product of the target user. Compared with the first embodiment, the embodiment receives the request. No need to extract user's purchase data from the business server, shorten the analysis and forecast time, and improve work efficiency.

In addition, the embodiment of the present application further provides a computer readable storage medium, where the computer readable storage medium includes a product purchase prediction program 10, and when the product purchase prediction program 10 is executed by the processor, the following operations are implemented:

Receiving step: receiving an analysis request with the target user identity information and the target product type;

The extracting step: extracting, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type;

Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.

Preferably, the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product includes the following training steps:

S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.

Preferably, the mean processing step further comprises:

The average number of occurrences of consecutive purchase behaviors in the purchase time series of the target product purchased by the preset number of users is averaged, and the average number of consecutive purchase behaviors of all users for various preset time intervals of the target product is obtained. The various preset time intervals of the product and the average number of occurrences of the corresponding consecutive purchase behaviors constitute an average purchase time sequence corresponding to the target product.

Preferably, the second analysis model is a Granger model, and the second analysis model corresponding to one product comprises the following training steps:

S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the second analysis model of the training by using each sample data in the verification set after the training is completed;

S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.

Optionally, the application further provides another product purchase prediction method, the method comprising:

Receiving step: receiving purchase data of various product types corresponding to all users from a plurality of predetermined service servers in real time;

Extracting step: after receiving the analysis request with the target user identity information and the target product type, extracting the purchase data of the various product types corresponding to the target user identity information and the preset number of users related to the target product type Purchase data;

Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.

The specific implementation manner of the computer readable storage medium of the present application is substantially the same as the specific implementation manner of the foregoing product purchase prediction method, and details are not described herein again.

The serial numbers of the embodiments of the present application are merely for the description, and do not represent the advantages and disadvantages of the embodiments.

Through the description of the above embodiments, those skilled in the art can clearly understand that the foregoing embodiment method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be through hardware, but in many cases, the former is better. Implementation. Based on such understanding, the technical solution of the present application, which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM as described above). , a disk, an optical disk, including a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the methods described in the various embodiments of the present application.

The above is only a preferred embodiment of the present application, and thus does not limit the scope of the patent application, and the equivalent structure or equivalent process transformation made by the specification and the drawings of the present application, or directly or indirectly applied to other related technical fields. The same is included in the scope of patent protection of this application.

Claims (20)

1. A product purchase prediction method is applied to a server, wherein the method comprises:

   Receiving step: receiving an analysis request with the target user identity information and the target product type;

   The extracting step: extracting, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type;

   Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

   a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

   Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

   a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

   a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.
2. The product purchase prediction method according to claim 1, wherein the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product comprises the following training steps:

   S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

   S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

   S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

   S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.
3. The product purchase prediction method according to claim 1, wherein the mean value processing step further comprises:

   The average number of occurrences of consecutive purchase behaviors in the purchase time series of the target product purchased by the preset number of users is averaged, and the average number of consecutive purchase behaviors of all users for various preset time intervals of the target product is obtained. The various preset time intervals of the product and the average number of occurrences of the corresponding consecutive purchase behaviors constitute an average purchase time sequence corresponding to the target product.
4. The product purchase prediction method according to claim 1, wherein the second analysis model is a Granger model, and the second analysis model corresponding to one product comprises the following training steps:

   S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

   S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

   S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

   S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the second analysis model of the training by using each sample data in the verification set after the training is completed;

   S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.
5. A product purchase prediction method, characterized in that the method comprises:

   Receiving step: receiving purchase data of various product types corresponding to all users from a plurality of predetermined service servers in real time;

   Extracting step: after receiving the analysis request with the target user identity information and the target product type, extracting the purchase data of the various product types corresponding to the target user identity information and the preset number of users related to the target product type Purchase data;

   Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

   a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

   Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

   a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

   a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.
6. The product purchase prediction method according to claim 5, wherein the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product comprises the following training steps:

   S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

   S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

   S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

   S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.
7. The product purchase prediction method according to claim 5, wherein the second analysis model is a Granger model, and the second analysis model corresponding to one product comprises the following training steps:

   S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

   S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

   S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

   S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the second analysis model of the training by using each sample data in the verification set after the training is completed;

   S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.
8. A server, comprising: a memory, a processor and a display, wherein the memory stores a product purchase prediction program, and the product purchase prediction program is executed by the processor, and the following steps can be implemented:

   Receiving step: receiving an analysis request with the target user identity information and the target product type;

   The extracting step: extracting, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type;

   Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

   a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

   Mean processing step: performing an average processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

   a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

   a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.
9. The server according to claim 8, wherein the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product comprises the following training steps:

   S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

   S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

   S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

   S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.
10. The server according to claim 8, wherein the mean processing step further comprises:

    The average number of occurrences of consecutive purchase behaviors in the purchase time series of the target product purchased by the preset number of users is averaged, and the average number of consecutive purchase behaviors of all users for various preset time intervals of the target product is obtained. The various preset time intervals of the product and the average number of occurrences of the corresponding consecutive purchase behaviors constitute an average purchase time sequence corresponding to the target product.
11. The server according to claim 8, wherein the second analysis model is a Granger causal model, and the second analysis model corresponding to one product comprises the following training steps:

    S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

    S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

    S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

    S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the trained second analysis model by using each sample data in the verification set after the training is completed;

    S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.
12. A server, comprising: a memory, a processor and a display, wherein the memory stores a product purchase prediction program, and the product purchase prediction program is executed by the processor, and the following steps can be implemented:

    Receiving step: receiving purchase data of various product types corresponding to all users from a plurality of predetermined service servers in real time;

    Extracting step: after receiving the analysis request with the target user identity information and the target product type, extracting the purchase data of the various product types corresponding to the target user identity information and the preset number of users related to the target product type Purchase data;

    Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

    a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

    Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

    a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

    a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.
13. The server according to claim 12, wherein the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product comprises the following training steps:

    S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

    S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

    S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

    S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.
14. The server according to claim 12, wherein the second analysis model is a Granger causal model, and the second analysis model corresponding to one product comprises the following training steps:

    S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

    S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

    S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

    S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the second analysis model of the training by using each sample data in the verification set after the training is completed;

    S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.
15. A computer readable storage medium, comprising: a product purchase prediction program, wherein when the system purchase prediction program is executed by a processor, the following steps can be implemented:

    Receiving step: receiving an analysis request with the target user identity information and the target product type;

    The extracting step: extracting, from a plurality of predetermined service servers, purchase data of various product types corresponding to the target user identity information and purchase data of a preset number of users regarding the target product type;

    Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

    a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

    Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

    a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

    a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.
16. The computer readable storage medium according to claim 15, wherein the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product comprises the following training steps:

    S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

    S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

    S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

    S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.
17. The computer readable storage medium according to claim 15, wherein the mean processing step further comprises:

    The average number of occurrences of consecutive purchase behaviors in the purchase time series of the target product purchased by the preset number of users is averaged, and the average number of consecutive purchase behaviors of all users for various preset time intervals of the target product is obtained. The various preset time intervals of the product and the average number of occurrences of the corresponding consecutive purchase behaviors constitute an average purchase time sequence corresponding to the target product.
18. The computer readable storage medium according to claim 15, wherein the second analysis model is a Granger causal model, and the second analysis model corresponding to one product comprises the following training steps:

    S21. Extract purchase data of various product types corresponding to a preset number of users from a plurality of predetermined service servers, and generate purchase time corresponding to each product for each user according to the corresponding purchase time point in the purchase data. sequence;

    S22: Perform user selection one by one from the preset number of users until all users select, and after selecting one user, determine, according to the mapping relationship between the product type and the pre-trained first analysis model, determine corresponding products purchased by the user. The first analysis model, the purchase time series corresponding to the various products purchased by the user are respectively input into the corresponding first analysis model, and the first purchase prediction values corresponding to the various products are generated, and the preset number of users are respectively respectively generated. The purchase time series of each product is subjected to mean processing, generating an average purchase time series corresponding to various products of the preset number of users, and inputting an average purchase time series corresponding to each product into the corresponding first analysis model Generating, according to the second purchase prediction value corresponding to each product, the first purchase predicted value corresponding to the first purchase of each product, the first purchase predicted value corresponding to the other product types other than the product, and the kind The second purchase predicted value corresponding to the product is used as sample data of the product of the user;

    S23. Divide the sample data into a training set of a first ratio and a verification set of a second ratio, where the first ratio is greater than the second ratio;

    S24. Train the second analysis model by using each sample data in the training set, and verify the accuracy of the second analysis model of the training by using each sample data in the verification set after the training is completed;

    S25. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, the number of samples is increased by increasing the number of users in S21. Then, step S22 is performed.
19. A computer readable storage medium, comprising: a memory, a processor, and a display, wherein the memory stores a product purchase prediction program, and the product purchase prediction program is used by the processor Execution, the following steps can be achieved:

    Receiving step: receiving purchase data of various product types corresponding to all users from a plurality of predetermined service servers in real time;

    Extracting step: after receiving the analysis request with the target user identity information and the target product type, extracting the purchase data of the various product types corresponding to the target user identity information and the preset number of users related to the target product type Purchase data;

    Generating step: generating a corresponding purchase time series (X, Y) according to the purchase time point of the extracted purchase data, wherein X represents the interval between consecutive purchases of the same number of days, and Y represents the continuous purchase behavior of the same number of days Number of occurrences;

    a first prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to each product purchased by the target user, and inputting a purchase time series corresponding to each product into the corresponding In the first analysis model, generating a first purchase prediction value corresponding to each product;

    Mean processing step: performing a mean processing on the purchase time series of the target product of the preset number of users, and generating an average purchase time sequence corresponding to the target product of the preset number of users;

    a second prediction step: determining, according to a mapping relationship between the product type and the pre-trained first analysis model, a first analysis model corresponding to the target product purchased by the preset number of users, and inputting the average purchase time series to the corresponding first In an analysis model, generating a second purchase predicted value of the target product;

    a final prediction step: determining, according to a mapping relationship between the product type and the pre-trained second analysis model, a second analysis model corresponding to the target product purchased by the target user, and the first purchase predicted value of the target product, in addition to the target product A second purchase model corresponding to the first purchase predicted value corresponding to the other product type of the type and the second purchase predicted value of the target product is input, and a final purchase predicted value of the target product is generated.
20. The computer readable storage medium according to claim 19, wherein the first analysis model is a long-term and short-term memory network model, and the first analysis model corresponding to one product comprises the following training steps:

    S11. Extracting, from a plurality of predetermined service servers, purchase data of a certain product type corresponding to a preset number of users, and generating, according to the purchase time point in the purchase data, a purchase time corresponding to the product for each user. sequence;

    S12. The purchase time series corresponding to the product is divided into a training set of a first ratio and a verification set of a second ratio, wherein the first ratio is greater than the second ratio;

    S13: training the first analysis model by using a purchase time series in the training set, and verifying the accuracy of the first analysis model by using a purchase time series in the verification set after the training;

    S14. If the accuracy is greater than the preset threshold, the training is completed. If the accuracy is less than or equal to the preset threshold, increase the number of users in S11, and increase the number of purchase time series, and then perform step S12.

Images