Product recommendation method and device
Determine potential users through the liquidity type and activity prediction model and push recommended products, solving the problem that traditional service models are difficult to meet the needs of diversified customers and improving customer quality and stickiness.
Patent Information
- Application Number
- CN202410408387.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-07
- Publication Date
- 2024-11-22
AI Technical Summary
Traditional service models are difficult to meet the diverse and differentiated needs of customers. Account managers have limited resources and lack precise push for target customers, resulting in difficulties in improving customer quality and stickiness.
By obtaining transaction data, entering liquidity type and activity prediction models, identifying potential users, and pushing recommended products based on purchase behavior, improving customer quality and stickiness.
It realizes accurate recommendations to customers, improves customer quality and stickiness, and increases the bank's dependence on high-value customers.
Smart Images

Figure CN119006086A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a product recommendation method and device. Background Art
[0002] With the development of digital economy and digital industry chain, the business management model of enterprise customers has changed. Customer needs are diversified, differentiated and integrated. The traditional service model is difficult to meet the "one-to-one" companion service for all customers. The resources of the account manager team are limited, and marketing services rely on traditional means. There is a lack of accurate push of target customers. It requires manual phone calls and text messages, which is time-consuming and labor-intensive, and the manual processing efficiency is low. It is difficult to fully tap and enhance the comprehensive value of customers. Summary of the invention
[0003] The main purpose of the embodiments of the present invention is to provide a product recommendation method and device to recommend suitable products to customers, which helps to improve the quality and stickiness of customers.
[0004] In order to achieve the above object, an embodiment of the present invention provides a product recommendation method, comprising:
[0005] Obtaining current transaction data of the first account under each business mode, inputting the current transaction data into a working capital type prediction model corresponding to each business mode, and obtaining working capital type data;
[0006] determining an amount slope change index according to the transaction amount of the current transaction data of the second account, and inputting the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data;
[0007] Determine potential users according to the liquidity type data and the activity data;
[0008] Recommended products are pushed to the potential user based on the products purchased by the potential user.
[0009] In one embodiment, it also includes:
[0010] Training a preset classification algorithm model according to the training data under each business mode of the first account to obtain a classification training model corresponding to each business mode;
[0011] Testing the corresponding classification training model according to the test data under each business mode of the first account to obtain the corresponding initial model success rate;
[0012] The corresponding classification training model is adjusted according to the success rate of the initial model to obtain the working capital type prediction model.
[0013] In one embodiment, adjusting the corresponding classification training model according to the success rate of the initial model to obtain the working capital type prediction model includes:
[0014] When the success rate of the initial model is greater than or equal to a preset success rate threshold, the classification training model is continuously trained according to the training data under each business mode of the first account to obtain the working capital type prediction model;
[0015] When the success rate of the initial model is less than a preset success rate threshold, the classification training model is abandoned.
[0016] In one embodiment, the classification training model is continuously trained according to the training data under each business mode of the first account to obtain the working capital type prediction model, including:
[0017] Continue to train the classification training model according to the training data under each business mode of the first account to obtain a classification training intermediate model;
[0018] Testing the corresponding classification training intermediate model according to the test data under each business mode of the first account to obtain the corresponding intermediate model success rate;
[0019] Determine the decreasing success rate data according to the intermediate model success rate and the initial model success rate;
[0020] When the decreasing success rate data is greater than the preset decreasing threshold, the working capital type prediction model is obtained according to the model corresponding to the maximum value of the intermediate model success rate and the initial model success rate, otherwise the classification training model continues to be trained.
[0021] In one embodiment, pushing recommended products to the potential user according to the products purchased by the potential user includes:
[0022] Determine a recommended product corresponding to the product purchased by the potential user according to a product rule set;
[0023] The recommended product is pushed to the potential user.
[0024] In one embodiment, it also includes:
[0025] Acquire product data, and calculate frequent item sets of the product data;
[0026] The product rule set is determined according to the frequent item sets of the product data.
[0027] In one embodiment, it also includes:
[0028] Get the current transaction data of the initial second account;
[0029] After encrypting the current transaction data of the initial second account, the data is sent to an external federated learning device, so that the federated learning device returns extended transaction data according to the encrypted current transaction data and the encrypted data of the participant;
[0030] The current transaction data of the initial second account is updated according to the extended transaction data to obtain the current transaction data of the second account.
[0031] The embodiment of the present invention further provides a product recommendation device, comprising:
[0032] A working capital type module, used to obtain current transaction data of the first account under each business mode, input the current transaction data into a working capital type prediction model corresponding to each business mode, and obtain working capital type data;
[0033] an activity prediction module, used to determine an amount slope change index according to the transaction amount of the current transaction data of the second account, and input the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data;
[0034] A potential user module, used to determine potential users according to the liquidity type data and the activity data;
[0035] The push module is used to push recommended products to the potential user based on the products purchased by the potential user.
[0036] In one embodiment, it also includes:
[0037] A classification training model module, used to train a preset classification algorithm model according to the training data under each business mode of the first account, to obtain a classification training model corresponding to each business mode;
[0038] An initial model success rate module, used to test the corresponding classification training model according to the test data under each business mode of the first account to obtain the corresponding initial model success rate;
[0039] The working capital type prediction model module is used to adjust the corresponding classification training model according to the success rate of the initial model to obtain the working capital type prediction model.
[0040] In one embodiment, the working capital type prediction model module includes:
[0041] A working capital type prediction model unit, configured to, when the success rate of the initial model is greater than or equal to a preset success rate threshold, continue to train the classification training model according to the training data under each business mode of the first account to obtain the working capital type prediction model;
[0042] The abandonment unit is used to abandon the classification training model when the success rate of the initial model is less than a preset success rate threshold.
[0043] In one embodiment, the working capital type prediction model unit includes:
[0044] The classification training intermediate model subunit is used to continue training the classification training model according to the training data under each business mode of the first account to obtain the classification training intermediate model;
[0045] The intermediate model success rate subunit is used to test the corresponding classification training intermediate model according to the test data under each business mode of the first account to obtain the corresponding intermediate model success rate;
[0046] A decreasing success rate subunit, used to determine decreasing success rate data according to the intermediate model success rate and the initial model success rate;
[0047] The working capital type prediction model subunit is used to obtain the working capital type prediction model according to the model corresponding to the maximum value of the intermediate model success rate and the initial model success rate when the decreasing success rate data is greater than the preset decreasing threshold, otherwise continue to train the classification training model.
[0048] In one embodiment, the push module includes:
[0049] A recommended product determination unit, configured to determine a recommended product corresponding to the product purchased by the potential user according to a product rule set;
[0050] A push unit is used to push the recommended product to the potential user.
[0051] In one embodiment, it also includes:
[0052] A frequent itemset module, used to obtain product data and calculate frequent itemsets of the product data;
[0053] The product rule set module is used to determine the product rule set according to the frequent item sets of the product data.
[0054] In one embodiment, it also includes:
[0055] A current transaction data acquisition module, used to acquire the current transaction data of the initial second account;
[0056] An encryption module, used for encrypting the current transaction data of the initial second account and sending the encrypted data to an external federated learning device, so that the federated learning device returns extended transaction data according to the encrypted current transaction data and the encrypted data of the participant;
[0057] An updating module is used to update the current transaction data of the initial second account according to the extended transaction data to obtain the current transaction data of the second account.
[0058] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the product recommendation method when executing the computer program.
[0059] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the product recommendation method are implemented.
[0060] An embodiment of the present invention further provides a computer program product, including a computer program / instruction, which implements the steps of the product recommendation method when the computer program / instruction is executed by a processor.
[0061] The product recommendation method and device of the embodiment of the present invention first inputs the transaction data of the first account into the working capital type prediction model corresponding to each business mode to obtain working capital type data, then determines the amount slope change index according to the transaction amount of the second account and inputs it into the activity prediction model together with the transaction data of the second account to obtain activity data, and then determines the potential user according to the working capital type data and the activity data, and can push recommended products to the potential user according to the products purchased by the potential user, which helps to improve the quality and stickiness of customers. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0063] Figure 1 is a flow chart of a product recommendation method in an embodiment of the present invention;
[0064] Figure 2 is a flow chart of obtaining a working capital type prediction model in an embodiment of the present invention;
[0065] Figure 3 is a flow chart of S203 in an embodiment of the present invention;
[0066] Figure 4 is a flow chart of S301 in an embodiment of the present invention;
[0067] Figure 5 is a flow chart of obtaining current transaction data of a second account in an embodiment of the present invention;
[0068] Figure 6 is a flow chart of S104 in an embodiment of the present invention;
[0069] Figure 7 is a flow chart of determining a product rule set in an embodiment of the present invention;
[0070] Figure 8 is a structural block diagram of a product recommendation device in an embodiment of the present invention;
[0071] Fig. 9 is a structural block diagram of a product recommendation device in another embodiment of the present invention;
[0072] Fig.10 is a structural block diagram of a data preprocessing device in an embodiment of the present invention;
[0073] Fig.11 It is a structural block diagram of a client quality improvement device in an embodiment of the present invention;
[0074] Fig.12 is a structural block diagram of an immobile user identification device according to an embodiment of the present invention;
[0075] Fig.13 is a structural block diagram of a product recommendation device in an embodiment of the present invention;
[0076] Fig.14 is a structural block diagram of an output device in an embodiment of the present invention;
[0077] Fig.15 A schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. DETAILED DESCRIPTION
[0078] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0079] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.
[0080] The acquisition, storage, use, and processing of data in the technical solution of the present invention are in compliance with the relevant provisions of national laws and regulations. The user information in the embodiments of this application is obtained through legal and compliant channels, and the acquisition, storage, use, and processing of user information are authorized and agreed by the customer.
[0081] In order to solve the problem of not improving customer quality in the field of customer quality management, and to combine customer information to make intelligent recommendations for financial products, the present invention mines high-quality customers through a naive Bayes algorithm, identifies potential corporate inactive accounts through a neural network algorithm, and implements product marketing for corporate customers through an FP-growth algorithm.
[0082] Customer quality improvement refers to finding high-quality and expandable customers among a large amount of customer information, providing data support for subsequent further financial management recommendation marketing. In addition, in the bank information, there are also some inactive accounts that store funds for a long time but have no capital flow. These customers are likely to lack marketing partners or investment opportunities. Customers who have been watching the market are also potential high-quality customers for banks. The discovery of product recommendations is to provide these two types of customers with more suitable banking products for them, which helps to increase customers' dependence on banks, thereby improving the quality and stickiness of the bank's overall customers (such as corporate customers).
[0083] The present invention mainly comprises the following steps:
[0084] Step 1: Process the data into three data sets as required. The first data set is used to determine the type of liquidity, the second data set is used to identify the activity, and the third data set is used for product recommendations.
[0085] Step 2: By establishing a feature set of customer information, the algorithm model is constructed using the unitized naive Bayes classification algorithm. Considering that there are differences in the characteristics of customers of different categories and that customers of the same category may have common characteristics, the data must retain the customer's business methods. Then, a basic vector is constructed for each customer's basic information, and the historical data set is iteratively trained until the success rate exceeds the threshold.
[0086] Step 3: By establishing a feature set of inactive household information, a neural network algorithm is used to form an algorithm model to identify high-quality inactive households to be activated. In addition, in the databases of different participants, the same customer may have customer information with different levels of activity. For example, if the customer is a long-term inactive household, but there is a strong active information collection and query action in the government-related system, or there is cooperation project or bidding information in the leading cooperative enterprise, etc., all of these can serve as data support for judging customer information. Considering factors such as data confidentiality, a vertical federated learning framework can be established to obtain data from each participant.
[0087] Step 4: Use the FP-growth algorithm to calculate the frequent item sets of funds purchased by current bank customers, and obtain the frequent items of potential financial products, which are used to push the recommendation results to customers and support customers to manually add or delete recommended financial products of interest.
[0088] Step 5: Output a list of high-quality customers, a list of recommended inactive customers to be activated, and a list of recommended products.
[0089] The present invention is described in detail below with reference to the accompanying drawings.
[0090] Figure 1 is a flow chart of the product recommendation method in an embodiment of the present invention. Figure 1 As shown, the product recommendation method includes:
[0091] S101: Acquire current transaction data of a first account under each business mode, input the current transaction data into a working capital type prediction model corresponding to each business mode, and obtain working capital type data.
[0092] Before executing S101, the data used to train the model also needs to be cleaned. In order to improve data quality, the D1 data set (including training data under each business mode of the first account) retains the main business mode of the customer, basic information of the customer, asset information, transaction behavior, product usage and other characteristics, and removes customer information with abnormal customer status, the last financial transaction date greater than 2 years, manual counter control, credit abnormality, business registration status not in operation, and revoked business registration. The D2 data set (data set for training the activity prediction model) retains the attribute of the last financial transaction greater than 2 years on the basis of the D1 data set, deletes customer information less than 2 years, and adds third-party encrypted data obtained through the federated learning framework. The D3 data set (product data set for generating frequent product items) retains the internal information of the entire bank and forms a new data set only for data attributes. The present invention also forms the cleaned data into a standardized data format and then outputs it in a regularized manner, transforming it into a modeling data source that can be modeled and identified.
[0093] Figure 2 : is a flow chart of obtaining a working capital type prediction model in an embodiment of the present invention. Figure 2 As shown, the product recommendation method also includes:
[0094] S201: Train a preset classification algorithm model according to the training data under each business mode of the first account to obtain a classification training model corresponding to each business mode.
[0095] Assuming that the number of features in the data set D1 is n, each customer data is composed of multiple features. A customer information T is constructed using the VSM method to construct a vector space model. If the feature appears, it is recorded as 1, and if it does not appear, it is recorded as 0, so that each customer information constitutes an n-dimensional space vector t = {t1, t2, ..., t n}.
[0096] S202: Testing the corresponding classification training model according to the test data under each business mode of the first account to obtain the corresponding initial model success rate.
[0097] S203: adjusting the corresponding classification training model according to the success rate of the initial model to obtain the working capital type prediction model.
[0098] Figure 3 This is a flow chart of S203 in an embodiment of the present invention. Figure 3 As shown, S203 includes:
[0099] S301: When the success rate of the initial model is greater than or equal to a preset success rate threshold, the classification training model is continuously trained according to the training data under each business mode of the first account to obtain the working capital type prediction model.
[0100] In specific implementation, a spiral classification success rate threshold can be set. Considering that the capital flow processing methods of enterprise accounts of different main types are different, the success rate training results of each business method may be different. In addition, there is a gap in each calculation result of the naive Bayes algorithm model. In order to obtain the maximum success rate of the algorithm, the present invention sets the algorithm success rate threshold to a spiral progressive method. The maximum success rate of each business method is recorded as s, and each s value starts from 0.5. If the final success rate is less than 0.5, it means that the model prediction result is less than 50%, the model availability is too low, and the classification training model needs to be abandoned.
[0101] Figure 4 is a flow chart of S301 in an embodiment of the present invention. Figure 4 As shown, S301 includes:
[0102] S401: Continue to train the classification training model according to the training data under each business mode of the first account to obtain a classification training intermediate model.
[0103] S402: Testing the corresponding classification training intermediate model according to the test data under each business mode of the first account to obtain the corresponding intermediate model success rate.
[0104] In specific implementation, the naive Bayes classification algorithm can be used to build a classification training model. Each customer information is constructed as a vector t = {t1, t2, ..., t n}, the main business method of each enterprise is set as C, and the working capital category of each customer is set as c i , including R1 (high assets and high liquidity), R2 (high assets and low liquidity), R3 (low assets and high liquidity), and R4 (low assets and low liquidity). Among them, the level of assets can be set through a threshold. In the present invention, the limit is 10,000 RMB. R2 and R4 can be determined as high-quality corporate customers to be mined (i.e., potential users). Classification is to calculate the conditional probability p(c i |t) that the vector t corresponding to the current main business method is determined to be the category ci, and the largest probability can be determined to belong to the category c i . Relying on Bayes' criterion, we can obtain Expand the vector t into independent features, then p(t|c i ) = p(t1|c i )p(t2|c i )p(t3|c i )... p(t n |c i ), that is, the product of the probabilities of each item appearing in the category c i . In addition, in the actual calculation process, since the product of probabilities is the multiplication of many very small numbers, it is easy to cause calculation overflow. Therefore, the ln logarithm method is used for optimization during the product calculation, and finally the value of p(c i |t) is obtained. If p(c1|t) > p(c2|t), it is considered that the category corresponding to the t vector belongs to c1. If p(c1|t) < p(c2|t), it is considered that the category corresponding to the t vector belongs to c2. Use the retained cross-validation method to train the classification model. Each time in the spiral loop, one-third of the marked data is taken out as the training set, and the remaining two-thirds is used as the test set, and the average success rate is calculated by iterative training.
[0105] S403: Determine the decreasing success rate data according to the success rate of the intermediate model and the success rate of the initial model.
[0106] S404: When the decreasing success rate data is greater than the preset decreasing threshold, obtain the working capital type prediction model according to the model corresponding to the maximum value of the success rate of the intermediate model and the success rate of the initial model, otherwise continue to train the classification training model.
[0107] Specifically, when implemented, the success rate of each iteration increases by 0.01. If the success rate of the algorithm after each spiral calculation increases, continue to calculate; if the success rate starts to decrease in a spiral manner, and the total decrease is greater than the preset decreasing threshold of 0.1, then use the model corresponding to the maximum value of the previous success rate as the final model.
[0108] S302: When the success rate of the initial model is less than the preset success rate threshold, abandon the classification training model.
[0109] S102: Determine an amount slope change index according to the transaction amount of the current transaction data of the second account, input the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data.
[0110] In specific implementation, transaction data can reflect the change of funds to a certain extent, but as a special data sample for neural network, the characteristics are not obvious and need to be reconstructed to calculate the slope of amount change. The amount of funds is obtained once every quarter (4 months), and the 5 months before and after are set as the statistical window period. The change of funds in 24 windows (24 months) is calculated. The slope of the amount change in the jth window (jth month) is the slope of the 5 windows before and after, that is, in, k j is the slope of the amount change in the jth window, f l is the amount value of the lth window. If the amount trend changes sharply, it is considered to be abnormal. The amount slope change indicator is T, then When k j <k j-1 , D(j)=1, otherwise D(j)=0.
[0111] When training the activity prediction model, the sample data consisting of the historical transaction data and the historical amount slope change index of the second account are randomly divided, and 20% are taken as test samples and 80% as training samples. The Keras library is used to establish a neural network, and the input nodes and output nodes of the LM (Levenberg-Marquardt) neural network are set. The number of hidden layer nodes is set to 10, and the Adam method is used for solving. Relu(x)=max(x,0) can be used as the activation function in the hidden layer, which can improve the accuracy of the model. Set the number of model training times to 1000 times. If the classification accuracy of the training results reaches more than 90%, the model training is successful by default, and the model is saved as the final activity prediction model.
[0112] Figure 5 FIG. 1 is a flow chart of obtaining the current transaction data of the second account in an embodiment of the present invention. Figure 5 As shown, the product recommendation method also includes:
[0113] S501: Acquire current transaction data of the initial second account.
[0114] S502: After encrypting the current transaction data of the initial second account, send it to an external federated learning device, so that the federated learning device returns extended transaction data according to the encrypted current transaction data and the encrypted data of the participant.
[0115] In addition to the company's internal bank data, the company's industry development policies, peer development capabilities, competitive relationship capabilities, and changes in the company's asset management shareholders will all have an impact on the probability of the company's default. Therefore, the present invention improves the accuracy of judging corporate defaults by increasing the data sharing capabilities of federated learning.
[0116] In the specific implementation, a vertical federated learning relationship model is established. The execution subject bank server of the present invention is participant A, and a certain cooperative unit is participant B. The enterprise users included in participant A are (a, b, c), and the enterprise users included in participant B are (a, b). It is assumed that the current transaction data of the initial loan account of participant A contains features (X1, X2, X3), and the transaction data of participant B contains features (X4, X5, X6). A vertical federated learning relationship between A and B is established to obtain the comprehensive features of users a and b.
[0117] 1. Perform encrypted sample alignment to prevent non-cross users from being exposed at the enterprise perception level.
[0118] 2. Perform model encryption training on samples:
[0119] ① A third-party collaborator sends a public key to A and B to encrypt the data to be transmitted.
[0120] ②A and B encrypt and interact with each other based on the public key based on the features related to themselves (for bank server A, the features related to itself are the current transaction data of the initial second account), and calculate their respective gradients and losses based on the encrypted data.
[0121] ③A and B add masks to the encrypted data, gradients, and losses and send them to third-party collaborators.
[0122] ④ The third-party collaborator decrypts the encrypted data (including the current transaction data, gradient and loss encrypted by bank server A, and the encrypted data, gradient and loss of participant B) to generate extended transaction data including the characteristics of both A and B, and sends it back to A and B respectively.
[0123] S503: Update the current transaction data of the initial second account according to the extended transaction data to obtain the current transaction data of the second account.
[0124] During specific implementation, the present invention removes the mask in the extended transaction data, and updates the current transaction data of the initial second account according to the extended transaction data after the mask is removed.
[0125] S103: Determine potential users according to the liquidity type data and the activity data.
[0126] Among them, when the liquidity type is R2 (high assets and low liquidity) and R4 (low assets and low liquidity), the corresponding customers can be determined as potential users, and users whose activity data is higher than the preset activity threshold are potential users.
[0127] S104: Pushing recommended products to the potential user based on the products purchased by the potential user.
[0128] Figure 6 is a flow chart of S104 in an embodiment of the present invention. Figure 6 As shown, S104 includes:
[0129] S601: Determine, according to a product rule set, a recommended product corresponding to the product purchased by the potential user.
[0130] Figure 7 FIG. 1 is a flow chart of determining a product rule set in an embodiment of the present invention. Figure 7 As shown, the product recommendation method also includes:
[0131] S701: Acquire product data, and calculate frequent item sets of the product data.
[0132] In one embodiment, before executing S701, the method further includes:
[0133] The support parameter n and confidence parameter m are set individually to control the number of frequent items and the accuracy of association rules that can be generated by the current fund product set. The default n=0.5, that is, the proportion of the number of records containing the item set in the data set to the entire data set is greater than or equal to 0.5; the default m=0.75, that is, the proportion of the number of records in the item set to the number of records of each item is greater than or equal to 0.75.
[0134] Table 1
[0135]
[0136] In the specific implementation, the frequent item sets of fund products in the D3 dataset are calculated. Generally, it can be considered that if an item set is a non-frequent item set, then all its supersets are also non-frequent. Therefore, when the number of items in the set is greater than 0, a list of item sets consisting of h items is constructed in a loop, and the item sets with support greater than or equal to n are retained and an item set list consisting of h+1 items is generated until the number of items is equal to 0. The initial value h = 1, and the frequent items of product applications in the D3 dataset are output. Table 1 is a schematic table of frequent sets of fund products. As shown in Table 1, the frequent item generation results of support 0.5 and 0.7 are shown respectively.
[0137] S702: Determine the product rule set according to the frequent item sets of the product data.
[0138] Table 2
[0139]
[0140] In specific implementation, the number of records in the item set is further calculated based on the obtained frequent item set to the number of records in each item, and the proportion is compared with the m value. If it is greater than or equal to the m value, it is assumed to be an association rule that meets the conditions. Table 2 is an example table of the fund product rule set, which shows the rule generation results of confidence levels 0.7 and 0.6 respectively. For example, {Fa}->{Fc} means that when {Fa} fund product appears, {Fc} product may also appear; this method can select the possibility that when the product purchased by the user is {Fa}, the customer may also purchase product {Fc}.
[0141] S602: Push the recommended product to the potential user.
[0142] The execution subject of the product recommendation method of the present invention can be a bank server. The present invention uses machine learning technology to establish a customer activity prediction and product recommendation combination model based on bank data, predicts customers who are easy to improve, easy to activate, and high-value, and intelligently recommends products with high purchase probability for predicted customers, improves the work efficiency of account managers in managing accounts, fully explores and improves the comprehensive value contribution of customers, and helps banks recommend financial products to high-quality customer groups and carry out marketing activities.
[0143] Figure 8 is a structural block diagram of a product recommendation device in an embodiment of the present invention. Figure 8 As shown, the product recommendation devices include:
[0144] A working capital type module, used to obtain current transaction data of the first account under each business mode, input the current transaction data into a working capital type prediction model corresponding to each business mode, and obtain working capital type data;
[0145] An activity prediction module, used to determine an amount slope change index according to the transaction amount of the current transaction data of the second account, and input the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data;
[0146] A potential user module, used to determine potential users according to the liquidity type data and the activity data;
[0147] The push module is used to push recommended products to the potential user based on the products purchased by the potential user.
[0148] In one embodiment, it also includes:
[0149] A classification training model module, used to train a preset classification algorithm model according to the training data under each business mode of the first account, to obtain a classification training model corresponding to each business mode;
[0150] An initial model success rate module, used to test the corresponding classification training model according to the test data under each business mode of the first account to obtain the corresponding initial model success rate;
[0151] The working capital type prediction model module is used to adjust the corresponding classification training model according to the success rate of the initial model to obtain the working capital type prediction model.
[0152] In one embodiment, the working capital type prediction model module includes:
[0153] a working capital type prediction model unit, configured to, when the success rate of the initial model is greater than or equal to a preset success rate threshold, continue to train the classification training model according to the training data under each business mode of the first account to obtain the working capital type prediction model;
[0154] The abandonment unit is used to abandon the classification training model when the success rate of the initial model is less than a preset success rate threshold.
[0155] In one embodiment, the working capital type prediction model unit includes:
[0156] The classification training intermediate model subunit is used to continue training the classification training model according to the training data under each business mode of the first account to obtain the classification training intermediate model;
[0157] The intermediate model success rate subunit is used to test the corresponding classification training intermediate model according to the test data under each business mode of the first account to obtain the corresponding intermediate model success rate;
[0158] A decreasing success rate subunit, used to determine decreasing success rate data according to the intermediate model success rate and the initial model success rate;
[0159] The working capital type prediction model subunit is used to obtain the working capital type prediction model according to the model corresponding to the maximum value of the intermediate model success rate and the initial model success rate when the decreasing success rate data is greater than the preset decreasing threshold, otherwise continue to train the classification training model.
[0160] In one embodiment, the push module includes:
[0161] A recommended product determination unit, configured to determine a recommended product corresponding to the product purchased by the potential user according to a product rule set;
[0162] A push unit is used to push the recommended product to the potential user.
[0163] In one embodiment, it also includes:
[0164] A frequent itemset module, used to obtain product data and calculate frequent itemsets of the product data;
[0165] The product rule set module is used to determine the product rule set according to the frequent item sets of the product data.
[0166] In one embodiment, it also includes:
[0167] A current transaction data acquisition module, used to acquire the current transaction data of the initial second account;
[0168] An encryption module, used for encrypting the current transaction data of the initial second account and sending the encrypted data to an external federated learning device, so that the federated learning device returns extended transaction data according to the encrypted current transaction data and the encrypted data of the participant;
[0169] An updating module is used to update the current transaction data of the initial second account according to the extended transaction data to obtain the current transaction data of the second account.
[0170] Fig. 9 FIG. 1 is a structural block diagram of a product recommendation device in another embodiment of the present invention. Fig. 9 As shown, in actual application, the product recommendation device includes a data preprocessing device, a customer quality improvement device, a static user identification device, a product recommendation device and an output device. The data preprocessing device is connected to the customer quality improvement device, the static user identification device and the product device, and the customer quality improvement device, the static user identification device and the product recommendation device are connected to the output device.
[0171] Data preprocessing device: preprocess the data set and parse it into three data sets D, D = {D1, D2, D3}, where the D1 data set is used for the customer quality improvement device, the D2 data set is used for the inactive user identification device, and the D3 data set is used for the product recommendation device.
[0172] Fig.10 is a structural block diagram of a data preprocessing device in an embodiment of the present invention. Fig.10 As shown, the data preprocessing device includes a data acquisition unit 11, a data cleaning unit 12, an attribute construction unit 13, and a data change unit 14, wherein:
[0173] Data acquisition unit 11: used to import original bank data sets.
[0174] Data cleaning unit 12: In order to improve data quality, the D1 dataset retains the main business mode of customers, basic customer information, asset information, transaction behavior, product usage and other characteristics, and removes customer information with abnormal customer status, the last financial transaction date greater than 2 years, manual counter control, credit abnormality, business registration status not in operation, and revoked business registration. Based on the D1 dataset, the D2 dataset retains the attribute of the last financial transaction greater than 2 years, deletes customer information less than 2 years, and adds third-party encrypted data obtained through the federated learning framework. The D3 dataset retains the internal information of the entire bank and forms a new dataset only for data attributes.
[0175] Attribute construction unit 13: forms the cleaned data into a standardized data format.
[0176] Data transformation unit 14: used to output the attribute-constructed data in a regularized manner and transform it into a modeling data source that can be modeled and identified.
[0177] Customer quality improvement device: Use the D1 data set to establish a feature set of customer information, use the naive Bayes classification algorithm to build an algorithm model, build a basic vector for each customer's basic information, and iteratively train the historical data set until the success rate exceeds the threshold.
[0178] Fig.11 is a structural block diagram of a customer quality improvement device in an embodiment of the present invention. Fig.11 As shown, the customer quality improvement device includes a threshold setting unit 21, a vector space construction unit 22, a classification calculation unit 23, and a model output unit 24, wherein:
[0179] Threshold setting unit 21: including a working capital type module, which is used to set the spiral classification success rate threshold. Considering that the capital flow processing methods of enterprise accounts of different main types are different, the success rate training results of each business method may be different. In addition, there is a gap in each calculation result of the naive Bayes algorithm model. In order to obtain the maximum success rate of the algorithm, the present invention sets the algorithm success rate threshold to a spiral progressive method. The maximum success rate of each business method is recorded as s, and the value of each s starts from 0.5. If the final success rate is less than 0.5, it means that the model prediction result is less than 50%, the model availability is too low, and the classification training model needs to be abandoned. Then the success rate of each iteration is progressive by 0.01. If the algorithm success rate after each spiral calculation is progressive, the calculation continues; if the success rate begins to decrease in a spiral, and the total decrease is greater than the preset decrease threshold 0.1, the model corresponding to the maximum value of the previous success rate is used as the final model.
[0180] Vector space construction unit 22: It includes a liquid capital type module. Assuming that the number of features in dataset D1 is n, each customer data consists of multiple features. A customer information T constructs a vector space model using the VSM method. If a feature appears, it is recorded as 1, and if it does not appear, it is recorded as 0, so that each customer information forms an n-dimensional space vector t = {t1, t2,..., t n}.
[0181] Classification calculation unit 23: It includes a classification training model module, an initial model success rate module, and a liquid capital type prediction model module. It is used to construct a classification training model using the Naive Bayes classification algorithm. Each customer information is constructed as a vector t = {t1, t2,..., t n}, the main business method of each enterprise is set as C, and the liquid capital category of each customer is set as c i , including R1 (high assets and high liquidity), R2 (high assets and low liquidity), R3 (low assets and high liquidity), and R4 (low assets and low liquidity). The level of assets can be set through a threshold. In the present invention, it is limited to 10,000 RMB. R2 and R4 can be determined as high-quality corporate customers that can be mined (i.e., potential users). Classification is to calculate the conditional probability p(c i |t) that the vector t corresponding to the current main business method is determined to be the category ci, and the largest probability can be determined to belong to the category c i . Depending on the Bayesian criterion, we can get Expand the vector t into independent features, then p(t|c i ) = p(t1|c i )p(t2|c i )p(t3|c i )... p(t n |c i ), that is, the product of the probabilities of each term appearing in the category c i . Additionally, in the actual calculation process, since the product of probabilities is the multiplication of many very small numbers, it is easy to cause calculation overflow. Therefore, the ln logarithm method is used for optimization during the product calculation, and finally the value of p(c i |t) is obtained. If p(c1|t) > p(c2|t), it is considered that the category corresponding to the t vector belongs to c1. If p(c1|t) < p(c2|t), it is considered that the category corresponding to the t vector belongs to c2. The classification model is trained using the retained cross-validation method. Each time in the spiral loop, one-third of the labeled data is taken out as the training set, and the remaining two-thirds is used as the test set. The average success rate is calculated through iterative training until the success rate reaches the set threshold of the current spiral, and then the data fetching and training start again until the success rate begins to decrease spirally. If the total decrease is greater than 0.1, the model corresponding to the maximum value of the previous success rate is used as the final model.
[0182] Model output unit 24: provides the above training model to the product recommendation device and output device.
[0183] Inactive household identification device: The D2 data set is trained through the LM neural network model to construct a neural network model, classify and predict the real-time diagnostic data, and identify the list of inactive households that can be recommended for activation.
[0184] Fig.12 is a structural block diagram of a static user identification device in an embodiment of the present invention. Fig.12 As shown, the immobile user identification device includes a federated learning unit 30, a slope calculation unit 31, a model training unit 32, and an identification output unit 33, wherein:
[0185] The federated learning unit 30 includes a current transaction data acquisition module, an encryption module and an update module.
[0186] Slope calculation unit 31: Transaction data can reflect the change of funds to a certain extent, but as a special data sample of the neural network, the characteristics are not obvious and need to be reconstructed to calculate the slope of the amount change. The faster the amount changes, the higher the customer activity is. The amount of funds is obtained once every quarter (4 months), and the 5 months before and after are set as the statistical window period. The change of funds in 24 windows (24 months) is calculated. The slope of the amount change in the jth window (jth month) is the slope of the 5 windows before and after, that is, in, k j is the slope of the amount change in the jth window, f l is the amount value of the lth window. If the amount trend changes sharply, it is considered to be abnormal. The amount slope change indicator is T, then When k j <k j-1 , D(j)=1, otherwise D(j)=0.
[0187] Model training unit 32: includes an activity prediction module, which is used to randomly divide the sample data composed of the historical transaction data and the historical amount slope change index of the second account, taking 20% as test samples and 80% as training samples. Use the Keras library to establish a neural network, set the input nodes and output nodes of the LM (Levenberg-Marquardt) neural network, set the number of hidden layer nodes to 10, and use the Adam method for solving. Relu(x)=max(x,0) can be used as the activation function in the hidden layer, which can improve the accuracy of the model. Set the number of model training times to 1000 times, and the classification accuracy of the training results reaches more than 90%. The model training is successful by default, and the model is saved as the final activity prediction model.
[0188] Identification output unit 33: inputs real-time data, enters it into the neural network model after slope calculation, and obtains the output node of the model as the judgment result. For accounts with high activity as the judgment result, the corresponding customer information of the account is transmitted to the product recommendation device and the result output device.
[0189] Product recommendation device: Use the FP-growth algorithm to calculate the frequent item sets of funds purchased by current bank customers at the D3 data level, and obtain the potential frequent items of financial products. Based on the high-quality customer groups obtained by the previous two devices, a list of recommended user products can be formed, and the recommended financial products results can be pushed to customers.
[0190] Fig.13 is a structural block diagram of a product recommendation device in an embodiment of the present invention. Fig.13 As shown, the product recommendation device includes a data screening unit 41, a parameter setting unit 42, a frequent item generating unit 43, an association rule output unit 44, and a recommendation result output unit 45.
[0191] Data screening unit 41: obtains D3 data source and only retains basic customer information and fund products.
[0192] Parameter setting unit 42: Personalized setting of support parameter n and confidence parameter m, used to control the number of frequent items that can be generated by the current fund product set and the accuracy of the association rules. The default value of n = 0.5, that is, the number of records containing item sets in the data set accounts for a proportion greater than or equal to 0.5 of the entire data set; the default value of m = 0.75, that is, the number of records in the item set accounts for a proportion greater than or equal to 0.75 of the number of records in each item.
[0193] Frequent item generation unit 43: includes a frequent item set module, which is used to calculate the frequent item sets of fund products in the D3 data set. It can be generally considered that if an item set is a non-frequent item set, then all its supersets are also non-frequent. Therefore, when the number of items in the set is greater than 0, a list of item sets consisting of h items is constructed in a loop, and item sets with support greater than or equal to n are retained and an item set list consisting of h+1 items is generated until the number of items is equal to 0, the initial value h=1, and the frequent items of product applications in the D3 data set are output.
[0194] The association rule output unit 44 includes a product rule set module, which further calculates the ratio of the number of records in the item set to the number of records in each item according to the obtained frequent item set, compares the ratio with the m value, and if it is greater than or equal to the m value, it is assumed to be an association rule that meets the conditions.
[0195] Recommendation result output unit 45: aggregates the recommended product set and outputs it to the output device.
[0196] Output device: output high-quality customer lists, recommended inactive customer lists to be activated, and product recommendation lists.
[0197] Fig.14 is a structural block diagram of an output device in an embodiment of the present invention. Fig.14 As shown, the output device includes: a high-quality corporate customer input unit 51, a potential inactive customer identification input unit 52, a recommended product input unit 53, and a result output unit 54.
[0198] High-quality corporate customer entry unit 51: includes a potential user module, supports entry of information on high-quality corporate customers (customers corresponding to the working capital types R2 (high assets and low liquidity) and R4 (low assets and low liquidity)), and generates a corresponding list.
[0199] Potential inactive user identification and entry unit 52: includes a potential user module, supports entry of potential inactive user information (user information whose activity data is higher than a preset activity threshold), and generates a corresponding list.
[0200] The recommended product input unit 53 includes a push module, supports inputting recommended product set information, and generates a corresponding list.
[0201] Result output unit 54: sorts and outputs all lists.
[0202] In summary, the product recommendation device of the embodiment of the present invention first inputs the transaction data of the first account into the working capital type prediction model corresponding to each business mode to obtain working capital type data, and then determines the amount slope change index according to the transaction amount of the second account and inputs it into the activity prediction model together with the transaction data of the second account to obtain activity data, and then determines the potential user according to the working capital type data and the activity data, and can push recommended products to the potential user according to the products purchased by the potential user, which helps to improve the quality and stickiness of customers.
[0203] Fig.15 FIG. 9 is a schematic block diagram of the system structure of the electronic device 9600 according to an embodiment of the present application. Fig.15 As shown, the electronic device 9600 may include a central processor 9100 and a memory 9140; the memory 9140 is coupled to the central processor 9100. It is worth noting that Fig.15 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.
[0204] In one embodiment, the product recommendation method function may be integrated into the central processing unit 9100. The central processing unit 9100 may be configured to perform the following control:
[0205] Obtaining current transaction data of the first account under each business mode, inputting the current transaction data into a working capital type prediction model corresponding to each business mode, and obtaining working capital type data;
[0206] determining an amount slope change index according to the transaction amount of the current transaction data of the second account, and inputting the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data;
[0207] Determine potential users according to the liquidity type data and the activity data;
[0208] Recommended products are pushed to the potential user based on the products purchased by the potential user.
[0209] From the above description, it can be seen that the product recommendation method provided in the present application first inputs the transaction data of the first account into the working capital type prediction model corresponding to each business mode to obtain working capital type data, and then determines the amount slope change index according to the transaction amount of the second account and inputs it together with the transaction data of the second account into the activity prediction model to obtain activity data, and then determines the potential users based on the working capital type data and the activity data, and can push recommended products to the potential users based on the products purchased by the potential users, which helps to improve the quality and stickiness of customers.
[0210] In another embodiment, the product recommendation device may be configured separately from the central processor 9100. For example, the product recommendation device may be configured as a chip connected to the central processor 9100, and the functions of the product recommendation method may be implemented under the control of the central processor.
[0211] like Fig.15 As shown, the electronic device 9600 may also include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily have to include Fig.15 In addition, the electronic device 9600 may also include Fig.15 For components not shown, reference may be made to the prior art.
[0212] like Fig.15 As shown, the central processing unit 9100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor device and / or logic device. The central processing unit 9100 receives input and controls the operation of various components of the electronic device 9600.
[0213] The memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 9100 may execute the program stored in the memory 9140 to implement information storage or processing, etc.
[0214] The input unit 9120 provides input to the central processing unit 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to provide power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.
[0215] The memory 9140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 9140 may also be some other type of device. The memory 9140 includes a buffer 9141 (sometimes referred to as a buffer memory). The memory 9140 may include an application / function storage unit 9142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 9600 through the central processor 9100.
[0216] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).
[0217] The communication module 9110 is a transmitter / receiver 9110 that sends and receives signals via an antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processor 9100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.
[0218] Based on different communication technologies, multiple communication modules 9110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module, etc. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, thereby realizing a common telecommunication function. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processor 9100, so that recording can be performed on the local machine through the microphone 9132, and the sound stored on the local machine can be played through the speaker 9131.
[0219] The embodiment of the present invention also provides a computer-readable storage medium capable of implementing all the steps of the product recommendation method in the above embodiment, where the execution subject is a server or a client. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, all the steps of the product recommendation method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:
[0220] Obtaining current transaction data of the first account under each business mode, inputting the current transaction data into a working capital type prediction model corresponding to each business mode, and obtaining working capital type data;
[0221] determining an amount slope change index according to the transaction amount of the current transaction data of the second account, and inputting the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data;
[0222] Determine potential users according to the liquidity type data and the activity data;
[0223] Recommended products are pushed to the potential user based on the products purchased by the potential user.
[0224] In summary, the computer-readable storage medium of the embodiment of the present invention first inputs the transaction data of the first account into the working capital type prediction model corresponding to each business mode to obtain working capital type data, and then determines the amount slope change index according to the transaction amount of the second account and inputs it into the activity prediction model together with the transaction data of the second account to obtain activity data, and then determines the potential user according to the working capital type data and the activity data, and can push recommended products to the potential user according to the products purchased by the potential user, which helps to improve the quality and stickiness of customers.
[0225] The embodiment of the present invention also provides a computer program product capable of implementing all the steps of the product recommendation method in the above embodiment, where the execution subject is a server or a client. The computer program product includes a computer program / instruction. When the computer program / instruction is executed by a processor, all the steps of the product recommendation method in the above embodiment are implemented. For example, when the processor executes the computer program, the following steps are implemented:
[0226] Obtaining current transaction data of the first account under each business mode, inputting the current transaction data into a working capital type prediction model corresponding to each business mode, and obtaining working capital type data;
[0227] determining an amount slope change index according to the transaction amount of the current transaction data of the second account, and inputting the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data;
[0228] Determine potential users according to the liquidity type data and the activity data;
[0229] Recommended products are pushed to the potential user based on the products purchased by the potential user.
[0230] In summary, the computer program product of the embodiment of the present invention first inputs the transaction data of the first account into the working capital type prediction model corresponding to each business mode to obtain working capital type data, and then determines the amount slope change index according to the transaction amount of the second account and inputs it into the activity prediction model together with the transaction data of the second account to obtain activity data, and then determines potential users based on the working capital type data and the activity data, and can push recommended products to potential users based on the products purchased by potential users, which helps to improve customer quality and stickiness.
[0231] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the hardware + program embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0232] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0233] Although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment).
[0234] Although the present specification embodiment provides the method operation steps as described in the embodiment or flow chart, more or less operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiment is only one way in the order of execution of many steps, and does not represent a unique execution order. When the device or terminal product in practice is executed, it can be executed in sequence or in parallel (such as a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment) according to the method shown in the embodiment or the accompanying drawings. The term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, product or equipment including a series of elements not only includes those elements, but also includes other elements not clearly listed, or also includes elements inherent to such process, method, product or equipment. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or equipment including the elements.
[0235] For the convenience of description, the above devices are described in various modules according to their functions. Of course, when implementing the embodiments of this specification, the functions of each module can be implemented in the same or more software and / or hardware, or the module implementing the same function can be implemented by a combination of multiple sub-modules or sub-units. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0236] Those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the controller can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.
[0237] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0238] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0239] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0240] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0241] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0242] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0243] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, the embodiments of this specification may take the form of complete hardware embodiments, complete software embodiments or embodiments combining software and hardware. Moreover, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0244] The various embodiments in this specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The embodiments of this specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.
[0245] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiments of this specification. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0246] The above is only an example of the embodiment of the present specification and is not intended to limit the embodiment of the present specification. For those skilled in the art, the embodiment of the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiment of the present specification shall be included in the scope of the claims of the embodiment of the present specification.
Claims
1. A product recommendation method, characterized in that: include: Obtaining current transaction data of the first account under each business mode, inputting the current transaction data into a working capital type prediction model corresponding to each business mode, and obtaining working capital type data; determining an amount slope change index according to the transaction amount of the current transaction data of the second account, and inputting the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data; Determine potential users according to the liquidity type data and the activity data; Recommended products are pushed to the potential user based on the products purchased by the potential user.
2. The product recommendation method according to claim 1, characterized in that: Also includes: Training a preset classification algorithm model according to the training data under each business mode of the first account to obtain a classification training model corresponding to each business mode; Testing the corresponding classification training model according to the test data under each business mode of the first account to obtain the corresponding initial model success rate; The corresponding classification training model is adjusted according to the success rate of the initial model to obtain the working capital type prediction model.
3. The product recommendation method according to claim 2, characterized in that: According to the success rate of the initial model, the corresponding classification training model is adjusted to obtain the working capital type prediction model, which includes: When the success rate of the initial model is greater than or equal to a preset success rate threshold, the classification training model is continuously trained according to the training data under each business mode of the first account to obtain the working capital type prediction model; When the success rate of the initial model is less than a preset success rate threshold, the classification training model is abandoned.
4. The product recommendation method according to claim 3, characterized in that: The classification training model is further trained according to the training data under each business mode of the first account to obtain the working capital type prediction model including: Continue to train the classification training model according to the training data under each business mode of the first account to obtain a classification training intermediate model; Testing the corresponding classification training intermediate model according to the test data under each business mode of the first account to obtain the corresponding intermediate model success rate; Determine the decreasing success rate data according to the intermediate model success rate and the initial model success rate; When the decreasing success rate data is greater than the preset decreasing threshold, the working capital type prediction model is obtained according to the model corresponding to the maximum value of the intermediate model success rate and the initial model success rate, otherwise the classification training model continues to be trained.
5. The product recommendation method according to claim 1, characterized in that: Pushing recommended products to the potential user according to the products purchased by the potential user includes: Determine a recommended product corresponding to the product purchased by the potential user according to a product rule set; The recommended product is pushed to the potential user.
6. The product recommendation method according to claim 5, characterized in that: Also includes: Acquire product data, and calculate frequent item sets of the product data; The product rule set is determined according to the frequent item sets of the product data.
7. The product recommendation method according to claim 1, characterized in that: Also includes: Get the current transaction data of the initial second account; After encrypting the current transaction data of the initial second account, the data is sent to an external federated learning device, so that the federated learning device returns extended transaction data according to the encrypted current transaction data and the encrypted data of the participant; The current transaction data of the initial second account is updated according to the extended transaction data to obtain the current transaction data of the second account.
8. A product recommendation device, characterized in that: include: A working capital type module, used to obtain current transaction data of the first account under each business mode, input the current transaction data into a working capital type prediction model corresponding to each business mode, and obtain working capital type data; An activity prediction module, used to determine an amount slope change index according to the transaction amount of the current transaction data of the second account, and input the current transaction data of the second account and the amount slope change index into an activity prediction model to obtain activity data; A potential user module, used to determine potential users according to the liquidity type data and the activity data; The push module is used to push recommended products to the potential user based on the products purchased by the potential user.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the product recommendation method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the product recommendation method according to any one of claims 1 to 7 are implemented.
