Customer behavior mining method and device, terminal equipment and computer storage medium

Abstract

The present application relates to the technical field of data processing, and discloses a mining method and device of customer behavior, a terminal device and a storage medium. The present application first acquires the history information of each loan product of each to-be-mined customer; then, aggregates the history information of each loan product with preset crowd characteristic data to divide each to-be-mined customer into multiple customer groups; and finally, mines the loan behavior of the multiple customer groups. The technical scheme of the present application can mine the loan behavior of customers from the time dimension and the product dimension, thereby helping to comprehensively understand the loan behavior of different customer groups, understand customer demand, and identify the loan behavior mode of customers, and is conducive to the subsequent activities such as accurate marketing and risk identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a customer behavior mining method and device, a terminal device, and a computer storage medium. BACKGROUND

[0002] At present, the financial institutions mainly use the questionnaire survey + expert analysis method and the neural network model to analyze and mine the customer lending behavior. However, the traditional questionnaire survey + expert analysis method is heavily dependent on the questionnaire design and the experience of experts, and the behavior analysis takes a long time, which is difficult to quickly adapt to the changing market environment. In addition, the neural network model for lending behavior mining can quickly adapt to complex and changing customer needs, but the model calculation process is not conducive to understanding the internal rules of customer lending behavior, resulting in the model prediction results being uninterpretable.

[0003] In summary, the existing methods for analyzing and mining the customer lending behavior have been difficult to meet the application requirements of the financial institutions for discovering the customer lending behavior for precise marketing and risk identification. SUMMARY

[0004] The main purpose of the present application is to provide a customer behavior mining method, device, terminal device, and computer storage medium, which aims to accurately analyze and mine the customer lending behavior.

[0005] To achieve the above purpose, the present application provides a customer behavior mining method, which comprises:

[0006] obtaining the lending product history information of each customer to be mined;

[0007] aggregating the lending product history information with preset population characteristic data to divide each customer to be mined into a plurality of customer groups;

[0008] mining the lending behavior of the plurality of customer groups.

[0009] In some embodiments, the mining of the lending behavior of the plurality of customer groups comprises:

[0010] determining the current lending product transaction of a target mining customer in a target customer group to generate a current transaction data set;

[0011] learning the current transaction data set according to a preset association rule to generate a frequent lending product set;

[0012] mining the interdependence between the lending products of the target mining customer based on the frequent lending product set.

[0013] In some embodiments, the step of mining the interdependence between the loan products of the target mining client based on the set of frequent loan products comprises:

[0014] generating association rules based on the set of frequent loan products;

[0015] determining the interdependence between the loan products of the target mining client according to target rules that meet a preset confidence condition among all generated association rules.

[0016] In some embodiments, the mining of the loan behaviors of the plurality of customer groups comprises:

[0017] determining loan sequence data of a target mining client in a target customer group among the plurality of customer groups;

[0018] determining a latest frequent sequence pattern according to the loan sequence data;

[0019] mining the chronological relationship of the loan behaviors of the target mining client based on the latest frequent sequence pattern.

[0020] In some embodiments, the step of determining the loan sequence data of the target mining client can comprise:

[0021] determining historical basic data of the target mining client;

[0022] generating the loan sequence data of the target mining client according to the historical basic data.

[0023] In some embodiments, the step of aggregating the loan product historical information with preset customer group characteristic data to divide each of the mining clients into a plurality of customer groups comprises:

[0024] generating a loan history matrix according to the loan product historical information;

[0025] performing Gaussian clustering processing on a customer historical loan behavior vector obtained by vectorizing the loan history matrix and preset customer group characteristic data to divide each of the mining clients into a plurality of customer groups.

[0026] In addition, to achieve the above object, the present application further provides a client behavior mining device, which comprises:

[0027] an information acquisition module configured to acquire loan product historical information of each of the mining clients;

[0028] a customer grouping module configured to aggregate the loan product historical information with preset customer group characteristic data to divide each of the mining clients into a plurality of customer groups;

[0029] a behavior mining module configured to mine the lending behaviors of the plurality of customer groups.

[0030] In some embodiments, the behavior mining module comprises:

[0031] a product dimension mining unit configured to determine a current lending product transaction of a target mining customer in a target customer group of the plurality of customer groups to generate a current transaction dataset, learn the current transaction dataset according to a preset association rule to generate a frequent lending product set, and mine the interdependence between the lending products of the target mining customer based on the frequent lending product set.

[0032] The product dimension mining unit comprises:

[0033] a rule generation subunit configured to generate an association rule based on the frequent lending product set.

[0034] a mining subunit configured to determine the interdependence between the lending products of the target mining customer according to a target rule meeting a preset confidence condition in all the generated association rules.

[0035] The behavior mining module further comprises:

[0036] a time dimension mining unit configured to determine lending sequence data of a target mining customer in a target customer group of the plurality of customer groups, determine a latest frequent sequence pattern according to the lending sequence data, and mine the sequence relationship of the lending behaviors of the target mining customer based on the latest frequent sequence pattern.

[0037] The time dimension mining unit is further configured to determine historical basic data of the target mining customer, and generate the lending sequence data of the target mining customer according to the historical basic data.

[0038] The customer grouping module comprises:

[0039] a matrix generation unit configured to generate a lending history matrix according to the lending product historical information.

[0040] a clustering processing unit configured to perform Gaussian clustering processing on a customer historical lending behavior vector obtained by vectorizing the lending history matrix and preset population feature data, so as to divide each of the to-be-mined customers into a plurality of customer groups.

[0041] In the present application, each functional module of the customer behavior mining device realizes the steps of the customer behavior mining method as described above.

[0042] Further, in order to achieve the above object, the present application also provides a terminal device, comprising a memory, a processor and a customer behavior mining program stored in the memory and executable on the processor, wherein the customer behavior mining program, when executed by the processor, implements the steps of the customer behavior mining method as described above.

[0043] Further, in order to achieve the above object, the present application also provides a computer storage medium, wherein the computer storage medium stores a customer behavior mining program, and the customer behavior mining program, when executed by a processor, implements the steps of the customer behavior mining method as described above.

[0044] Further, in order to achieve the above object, the present application also provides a computer program product, comprising a computer program, wherein the computer program, when executed by a processor, implements the steps of the customer behavior mining method as described above.

[0045] The customer behavior mining method, device, terminal device, computer storage medium and computer program product provided by the present application first acquire the loan product historical information of each customer to be mined, then aggregate the loan product historical information with preset population characteristic data to divide the customers to be mined into multiple customer groups, and finally mine the loan behavior of the multiple customer groups.

[0046] In the process of mining the loan behavior of customers, the present application acquires the loan product historical information of multiple customers, aggregates the loan product historical information with population characteristic data to obtain multiple customer groups by grouping the multiple customers, and finally mines the loan behavior of the multiple customer groups. Thus, compared with the traditional customer behavior mining method using questionnaires and expert analysis or using a neural network model for prediction, the present application can comprehensively understand the loan behavior and customer demand of different customer groups by aggregating the historical loan product behavior of customers and population characteristic data, mining the loan behavior of different customer groups, accurately analyzing and mining the loan behavior of customers, and facilitating the subsequent activities such as precise marketing and risk identification of financial institutions. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The device structure diagram of the hardware running environment of the terminal device involved in the embodiment of the present application;

[0048] Figure 2 The flowchart of the first embodiment of the customer behavior mining method of the present application;

[0049] Figure 3An application flowchart involved in the mining method of customer behavior of the present application;

[0050] Figure 4 Another application flowchart involved in the mining method of customer behavior of the present application;

[0051] Figure 5 A function module diagram of the mining device of customer behavior of the present application.

[0052] The implementation, function features and advantages of the present application will be further explained with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0053] It should be understood that the specific embodiments described herein merely serve to explain the present application and do not limit the present application.

[0054] Reference Figure 1 , Figure 1 A device structure diagram of a hardware running environment of a terminal device involved in the embodiment scheme of the present application.

[0055] The terminal device of the embodiment of the present application can be a smart phone, a PC (Personal Computer), a tablet computer, a portable computer, a server and the like.

[0056] As shown in Figure 1 , the terminal device can include a processor 1001, for example, a CPU, a communication bus 1002, a user interface 1003, a network interface 1004 and a memory 1005. The communication bus 1002 is used to realize the connection communication between the components. The user interface 1003 can include a display screen (Display) and an input unit such as a keyboard (Keyboard). The optional user interface 1003 can further include a standard wired interface and a wireless interface. The network interface 1004 can optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface). The memory 1005 can be a high-speed RAM memory or a stable memory (non-volatile memory) such as a magnetic disk memory. The memory 1005 can optionally be a storage device independent of the aforementioned processor 1001.

[0057] Those skilled in the art can understand that Figure 1 the terminal device structure shown in the foregoing embodiments does not constitute a limitation on the terminal device, and can include more or fewer components than the diagram, or combine certain components, or different component arrangements.

[0058] As shown in Figure 1As shown, the memory 1005 as a computer storage medium can include an operating system, a network communication module, a user interface module, and a customer behavior mining program.

[0059] In Figure 1 In the terminal shown, the network interface 1004 is mainly used for connecting a background server and communicating data with the background server; the user interface 1003 is mainly used for connecting a client and communicating data with the client; and the processor 1001 can be used for calling the customer behavior mining program stored in the memory 1005 and performing the following operation steps:

[0060] Obtaining the historical information of the loan products of each customer to be mined;

[0061] Aggregating the historical information of the loan products with preset population characteristic data to divide each of the customers to be mined into a plurality of customer groups;

[0062] Mining the loan behaviors of the plurality of customer groups.

[0063] Further, the processor 1001 can also be used for calling the customer behavior mining program stored in the memory 1005 and performing the following operation steps:

[0064] In a target customer group of the plurality of customer groups, determining the current loan product transaction of a target mining customer to generate a current transaction data set;

[0065] Learning the current transaction data set according to a preset association rule to generate a frequent loan product set;

[0066] Mining the interdependence between the loan products of the target mining customer based on the frequent loan product set.

[0067] Further, the processor 1001 can also be used for calling the customer behavior mining program stored in the memory 1005 and performing the following operation steps:

[0068] Generating an association rule based on the frequent loan product set;

[0069] Determining the interdependence between the loan products of the target mining customer according to a target rule meeting a preset confidence condition in all the generated association rules.

[0070] Further, the processor 1001 can also be used for calling the customer behavior mining program stored in the memory 1005 and performing the following operation steps:

[0071] In a target customer group of the plurality of customer groups, determining the loan sequence data of a target mining customer;

[0072] determine a latest frequent sequential pattern according to the loan sequence data;

[0073] mine the loan behavior sequence of the target mining customer based on the latest frequent sequential pattern.

[0074] Further, the processor 1001 can also be configured to invoke the mining program of customer behavior stored in the storage 1005, and further perform the following operation steps:

[0075] determine the historical basic data of the target mining customer;

[0076] generate the loan sequence data of the target mining customer according to the historical basic data.

[0077] Further, the processor 1001 can also be configured to invoke the mining program of customer behavior stored in the storage 1005, and further perform the following operation steps:

[0078] generate a loan history matrix according to the loan product historical information;

[0079] perform Gaussian clustering processing on the customer historical loan behavior vector obtained by vectorizing the loan history matrix and preset population feature data, so as to divide the target mining customers into a plurality of customer groups.

[0080] Based on the above hardware structure, embodiments of the mining method of customer behavior are provided.

[0081] It should be noted that discovering the loan behavior pattern of the customer is of great significance for the bank to identify the user's loan intention for precise marketing and predict the customer's repayment default intention. Based on different business scenarios and application purposes, different institutions have different needs for mining customer loan behavior. At present, it is mainly divided into two categories. One is to understand and predict the customer's loan behavior pattern through traditional questionnaire survey-expert analysis method, that is, to analyze through small data + expert experience; the other is to predict the customer's loan risk through neural network model based on Internet loan behavior.

[0082] However, the traditional questionnaire survey-expert analysis method is heavily dependent on questionnaire design and expert experience, and it takes a long time, which is difficult to quickly adapt to the changing market environment. The experience-oriented scheme in the past often has limitations and cannot identify emerging loan habits in time. The loan risk mining based on neural network model can quickly adapt to complex and changing customer needs, but the model process is a black box, which is not conducive to understanding the internal rules of customer loan behavior, and has no explainability.

[0083] In summary, the existing way of analyzing and mining the loan behavior of customers has been difficult to meet the application needs of financial institutions to discover the loan behavior of customers for precise marketing and risk identification.

[0084] In view of the above phenomenon, the present application provides a customer behavior mining method, which considers customer historical lending product behavior and current lending situation, aggregates customer historical lending product behavior and crowd feature clustering to obtain different customer groups, thereby mining the sequence relationship of lending products based on customer lending sequence and mining the contemporaneous correlation of lending products based on current point customer lending behavior, and finally obtaining the lending behavior mode of different customer groups. That is, the present application mines customer lending behavior from the time dimension and product dimension, which helps to comprehensively understand the lending behavior of different customer groups, understand customer demand, identify customer lending behavior mode, and is beneficial to subsequent activities such as precise marketing and risk identification.

[0085] That is, in the process of mining the lending behavior of customers, the present application obtains the historical information of lending products of a plurality of customers, then aggregates the historical information of lending products with crowd feature data to process the plurality of customers to obtain a plurality of customer groups, and finally mines the lending behavior of the plurality of customer groups. In this way, compared with the traditional customer behavior mining method using questionnaires + expert analysis or using a neural network model for prediction, the present application aggregates customer historical lending product behavior and crowd feature data to obtain different customer groups, and then mines the lending behavior of different customer groups, which can comprehensively understand the lending behavior and customer demand of different customer groups, thereby achieving the purpose of accurately analyzing and mining customer lending behavior, and being beneficial to the subsequent activities such as precise marketing and risk identification of financial institutions.

[0086] Please refer to Figure 2 , Figure 2 The flowchart of the first embodiment of the customer behavior mining method of the present application. It should be noted that although the logical order is shown in the flowchart, in some cases, the customer behavior mining method of the present application can also execute the steps shown or described in a different order than here.

[0087] In the present embodiment, the customer behavior mining method of the present application is applied to the terminal device described above. The customer behavior mining method of the present application comprises:

[0088] Step S10, obtaining the historical information of lending products of each of the customers to be mined;

[0089] In the present embodiment, in the process of mining the lending behavior of one or more customers by the terminal device, the one or more customers are first taken as customers to be mined whose lending product behavior needs to be mined, and then the historical information of lending products of each of the customers to be mined is obtained from a pre-connected financial data platform or other alternative data platform.

[0090] S20, aggregate the loan product history information with preset crowd characteristic data to divide the to-be-mined customers into multiple customer groups;

[0091] In this embodiment, after obtaining the loan product history information of each to-be-mined customer, the terminal device immediately aggregates the loan product history information with preset one or more crowd characteristic data, thereby dividing the one or more to-be-mined customers into multiple customer groups.

[0092] It should be noted that in this embodiment, the same to-be-mined customer can belong to different customer groups.

[0093] Further, in a feasible embodiment, the above-mentioned step S20 can specifically include:

[0094] S201, generate a loan history matrix according to the loan product history information;

[0095] It should be noted that in this embodiment, the loan product history information includes whether each type of loan product is borrowed during the period.

[0096] In this embodiment, after obtaining the loan product history information of each to-be-mined customer, the terminal device further constructs a two-dimensional matrix respectively based on the loan product history information of each to-be-mined customer, and names the two-dimensional matrix as a loan history matrix.

[0097] For example, after obtaining the loan product history information of each to-be-mined customer, the terminal device constructs a two-dimensional matrix based on whether each type of loan product is borrowed during the period included in the information, by letting the x-axis direction represent the loan product type, the y-axis direction represent each period, and the specific value "yes" or "no" represent whether the loan product is borrowed, and then the terminal further names the two-dimensional matrix as a loan history matrix.

[0098] For example, assuming that the loan history matrix constructed by the terminal device based on the loan product history information of the to-be-mined customer A is: Then the matrix indicates that the to-be-mined customer A borrowed product i2 in the first period, borrowed product i1 in the second period, and borrowed products i1 and i3 in the third period.

[0099] S202, perform Gaussian clustering processing on the customer history loan behavior vector obtained by vectorizing the loan history matrix and the preset crowd characteristic data, to divide the to-be-mined customers into multiple customer groups.

[0100] In this embodiment, after the terminal device constructs a corresponding loan history matrix based on each customer to be mined, it further vectorizes the loan history matrix to obtain a customer's historical loan behavior vector. Furthermore, the terminal device performs Gaussian clustering processing on the customer's historical loan behavior vector and one or more preset population feature data, thereby dividing the customer to be mined into multiple customer groups.

[0101] It should be noted that, in this embodiment, the preset population characteristic data includes: occupation, marital status, workplace, etc.

[0102] For example, such as Figure 3 The customer segmentation process shown in this embodiment involves the terminal device vectorizing the historical lending behavior vector obtained by constructing a historical lending matrix based on the historical lending product information of the customer to be mined, and performing Gaussian clustering on the customer's demographic characteristics such as occupation, marital status, and workplace, thereby dividing the customer to be mined into multiple customer groups.

[0103] It should be noted that in this embodiment, since Gaussian clustering uses the mean and standard deviation, the final output of each sample is a probability. Therefore, a customer to be explored can belong to multiple different customer groups. This clustering method is closer to real life and conforms to the fact that a customer can belong to multiple product interest groups.

[0104] Furthermore, in this embodiment, Gaussian clustering performs clustering at each value of K according to a set range of the number of clusters K, and calculates the silhouette coefficients for different numbers of clusters K, ultimately calculating the silhouette coefficients of the clustering results. By plotting the silhouette coefficients as a line graph with respect to K, the K value with the largest silhouette coefficient is selected as the final number of clusters, thus obtaining K customer groups with duplicate customer appearances.

[0105] Step S30: Mining the lending behavior of multiple customer groups.

[0106] In this embodiment, after the terminal device obtains different customer groups by aggregating the historical information of the loan products of the customers to be mined and the demographic data, it mines the sequential relationship of loan products based on the customer's loan sequence and the synchronous correlation of loan products based on the customer's loan behavior at the current time, and finally obtains the loan behavior patterns of different customer groups.

[0107] Furthermore, in a feasible embodiment, step S30 described above may specifically include:

[0108] Step S301: Among the target customer groups in the multiple customer groups, determine the current loan product transactions of the target mining customers to generate a current transaction dataset.

[0109] In the embodiment, the terminal device determines a target customer group from the plurality of customer groups in turn when performing the loan behavior mining on the plurality of customer groups, and then determines a target mining customer from the plurality of to-be-mined customers in the target customer group, and generates a current node product transaction corresponding to a current transaction dataset of the target mining customer.

[0110] As shown in the customer loan behavior mining process, Figure 4 In the embodiment, the terminal device determines a target customer group X from the plurality of customer groups when performing the loan behavior mining, and then takes each to-be-mined customer in the customer group X as a target mining customer in turn to obtain the loan product data of the target mining customer in the current period from the product dimension, and forms a customer loan product transaction set T. That is, for a customer C in the customer group X, the current loan product transaction I can be expressed as: I = {i1, i2, i3,..., i m} Each i k (k = 1, 2, 3,..., m) represents a loan product of the current customer C, and if there are m customers, the current transaction dataset D = {I1, I2,..., Im} is generated. n

[0111] In step S302, the current transaction dataset is learned according to a preset association rule to generate a frequent loan product set.

[0112] In the embodiment, the terminal device further performs association rule analysis on the current transaction dataset after generating the current transaction dataset when performing the loan behavior mining on the target customer group, that is, the terminal device learns the dataset based on the pre-set association rule to generate a corresponding frequent loan product set.

[0113] It should be noted that, in the embodiment, the preset association rule is a mining rule for indicating the mutual dependence of loan products, which includes support and confidence.

[0114] As shown in the customer loan behavior mining process, Figure 4 The terminal device sets the support and confidence of the association rule, where the support represents the frequency of the item set in the transaction, and the confidence represents the credibility of the association rule. In this way, the confidence of i1=>i2 represents the possibility of the transaction containing i1 also containing i2 in the transaction dataset D. Based on this, the terminal device can obtain the frequent item set in the transaction dataset D, that is, the frequent loan product set, by learning the transaction dataset D based on the association rule.

[0115] In step S303, the mutual dependence between the loan products of the target mining customer is mined based on the frequent loan product set.

[0116] ​In the embodiment, the terminal device learns the corresponding frequent lending and borrowing product set from the current transaction data set according to the pre-set association rule, and further mines the interdependence between the lending and borrowing products of the target mining customer based on the frequent lending and borrowing product set.

[0117] Further, in a possible implementation, the step S303 can specifically include:

[0118] In step S3031, association rules are generated based on the frequent lending and borrowing product set.

[0119] In step S3032, the interdependence between the lending and borrowing products of the target mining customer is determined according to the target rules that meet the pre-set confidence condition from all the generated association rules.

[0120] For example, as shown in the figure, Figure 4 In the embodiment, the terminal device further mines the interdependence between the lending and borrowing products based on the frequent lending and borrowing product set after the association rule analysis, that is, the terminal device first finds all the non-empty subsets of the maximum frequent lending and borrowing product set, then generates all the possible association rules based on all the non-empty subsets, and finally calculates the confidence of all the rules in sequence to find all the rules R that are greater than the confidence threshold, wherein the rule R represents the interdependence between the lending and borrowing products, for example, {i1, i4} => i1 indicates that the customer who has borrowed i1 and i4 products is most likely to borrow i1 product.

[0121] Further, in another possible implementation, the step S30 can specifically further include:

[0122] In step S304, the lending and borrowing sequence data of the target mining customer is determined in the target customer group from the plurality of customer groups.

[0123] In the embodiment, when the terminal device mines the lending and borrowing behavior of the plurality of customer groups, the target customer group is determined in sequence from the plurality of customer groups, and then the lending and borrowing sequence data of the target mining customer that needs to be mined is determined from the plurality of mining customers in the target customer group.

[0124] Further, in a possible implementation, the step S304 can specifically include:

[0125] In step S3041, the historical basic data of the target mining customer is determined.

[0126] In step S3042, the lending and borrowing sequence data of the target mining customer is generated according to the historical basic data.

[0127] For example, as shown in the figure, Figure 4The shown customer loan behavior mining process, in this embodiment, the terminal device determines the loan sequence data of the target mining customer, first acquires the loan product history data of the target mining customer, and then forms the historical basic data of the target mining customer including customer ID, loan product time, and loan product code. Then, the terminal device sets the statistical period of the loan product, thereby merging the records with the same customer ID (such as monthly statistics, the same month loan product is in the same set), and arranging according to the loan time sequence, that is, the loan sequence data S of the target mining customer can be obtained.

[0128] For example, it is assumed that the terminal device determines the loan sequence data of each to-be-mined customer in the current target customer group as shown in the following table:

[0129]

[0130]

[0131] Among them, the loan sequence <{i1, i6}, {i2}> indicates that the customer C001 has loaned i1, i6 products in a period, and has loaned i2 products in the subsequent period.

[0132] Step S305, determining the latest frequent sequence pattern according to the loan sequence data;

[0133] In this embodiment, after the terminal device determines the loan sequence data of the target mining customer, the SPADE algorithm is further used to mine the loan sequence data to determine the latest frequent sequence pattern.

[0134] Exemplarily, as shown in the figure, Figure 4 The terminal device mines the loan sequence data of the target mining customer by using the SPADE algorithm, wherein the SPADE algorithm uses the combination property to decompose the original problem into sub-problems that can be solved in the main memory, and uses the search method and connection operation based on the sequence grid. The specific process is as follows: (1) generate all 1-sequences, the sequence composed of each single item is a 1-sequence, and count <{i1}><{i2}>...<{i m}>The sequence support corresponding to each sequence is greater than the support threshold. The sequence is a frequent 1-sequence, k=1. (2) The candidate sequence of 2-sequence is generated by the self-join operation of the frequent 1-sequence, and then the frequent 2-sequence is obtained, k=2. The time order of the borrowing behavior should be considered when associating. (3) The candidate sequence of 3-sequence is generated by the self-join operation of the frequent 2-sequence. There are three possible ways to self-join the 2-frequent itemsets: one is self-join of the itemset atomic items, the second is to join the itemset atomic items with the sequence atomic items, and the third is to self-join the sequence atomic items. In order to improve the computational efficiency, the SPADE algorithm divides the sequence into blocks by equivalence class. Equivalence class refers to sequences with the same prefix. (4) k increases, and the candidate k-sequence S is generated by self-joining the frequent k-1 sequence as the atomic item and the equivalence class as the basis. k If k-sequence S k If the set is empty, the operation stops, and the frequent subsequence found by k-1 is the latest frequent sequence pattern.

[0135] Step S306: Based on the latest frequent sequence pattern, mine the sequential relationship of the borrowing and lending behavior of the target customer.

[0136] For example, such as Figure 4 As shown, after determining the latest frequent sequence pattern, the terminal device can further mine the sequential relationship between lending and borrowing behaviors based on this latest frequent sequence pattern. That is, the terminal device first generates all possible sequence combinations based on the latest frequent sequence pattern; then, it calculates the confidence level of each sequence combination in turn, and finds all rules RS that are greater than the confidence level threshold. The rule RS represents the sequential relationship between lending and borrowing behaviors. For example, {i5}→{i7} means that a customer who borrowed product i5 is likely to borrow product i7 in the future.

[0137] This reveals the future borrowing intentions of different customer groups and the interdependence between borrowing products. Obtaining information on the preferences of different customer groups helps in understanding customer borrowing behavior and assists in targeted marketing and risk identification.

[0138] In the embodiment, the mining method of customer behavior provided by the embodiment of the application comprises the following steps: in the process of mining the lending behavior of one or more customers by a terminal device, first, the one or more customers are taken as to-be-mined customers who need to be mined for lending product behavior at present, then, the terminal device obtains the lending product historical information of each to-be-mined customer from a pre-connected financial data platform or other alternative data platform, and then, the terminal device immediately aggregates the lending product historical information and preset one or more customer group feature data, so as to divide the one or more to-be-mined customers into a plurality of customer groups. Finally, the terminal device mines the chronological relationship of the lending product based on the customer lending sequence and mines the contemporaneous correlation of the lending product based on the current time point customer lending behavior, and finally obtains the lending behavior mode of different customer groups.

[0139] In this way, compared with the traditional customer behavior mining method using questionnaires + expert analysis or using a neural network model for prediction, the application combines customer lending product history and customer group features to divide the groups, mines the lending intention, the chronological relationship of the lending behavior of different customer groups from the time dimension and the product dimension, and can automatically obtain the lending preferences of different customer groups.

[0140] In addition, the application mines the chronological relationship of the lending product based on the customer lending sequence and mines the contemporaneous correlation of the lending product based on the current time point customer lending behavior, and finally obtains the lending behavior mode of different customer groups. That is, in the technical solution of the application, the division of the customer groups and the mining of the lending behavior of the customers both consider the time dimension change of the customers, so that the application can fully identify the multi-dimensional lending behavior mode of the customers, and the mining result is more accurate.

[0141] In addition, the application further provides a mining device for customer behavior.

[0142] Please refer to Figure 5 , Figure 5 for the function module schematic diagram of an embodiment of the mining device for customer behavior of the application.

[0143] As Figure 5 shown, the mining device for customer behavior of the application comprises:

[0144] An information acquisition module 10 is configured to acquire the lending product historical information of each to-be-mined customer;

[0145] A customer group division module 20 is configured to aggregate the lending product historical information and preset customer group feature data to divide each to-be-mined customer into a plurality of customer groups;

[0146] A behavior mining module 30 is configured to mine the lending behavior of the plurality of customer groups.

[0147] Further, the behavior mining module comprises:

[0148] The product dimension mining unit is configured to determine current transaction data of a target mining customer in a target customer group among the plurality of customer groups, learn the current transaction data according to a preset association rule to generate a frequent borrowing product set, and mine interdependence between borrowing products of the target mining customer based on the frequent borrowing product set.

[0149] Further, the product dimension mining unit comprises:

[0150] A rule generation subunit is configured to generate association rules based on the frequent borrowing product set.

[0151] A mining subunit is configured to determine the interdependence between borrowing products of the target mining customer according to target rules that meet a preset confidence condition among all generated association rules.

[0152] Further, the behavior mining module further comprises:

[0153] A time dimension mining unit is configured to determine borrowing sequence data of a target mining customer in a target customer group among the plurality of customer groups, determine a latest frequent sequence pattern according to the borrowing sequence data, and mine a borrowing behavior sequence of the target mining customer based on the latest frequent sequence pattern.

[0154] Further, the time dimension mining unit is further configured to determine historical basic data of the target mining customer, and generate the borrowing sequence data of the target mining customer according to the historical basic data.

[0155] Further, the customer grouping module comprises:

[0156] A matrix generation unit is configured to generate a borrowing history matrix according to the borrowing product historical information.

[0157] A clustering processing unit is configured to perform Gaussian clustering processing on a customer historical borrowing behavior vector obtained by vectorizing the borrowing history matrix and preset population feature data, so as to divide the plurality of mining customers into a plurality of customer groups.

[0158] The functions of each module of the customer behavior mining device correspond to the steps in the customer behavior mining method, and the functions and implementation processes will not be described here.

[0159] The application further provides a computer storage medium, which stores a customer behavior mining program. When the customer behavior mining program is executed by a processor, the steps of the customer behavior mining method according to any one of the above embodiments are implemented.

[0160] The specific embodiment of the computer storage medium of the present application is basically the same as the above-mentioned mining method of customer behavior, and will not be repeated here.

[0161] The present application also provides a computer program product, which comprises a computer program, and the computer program realizes the steps of the mining method of customer behavior according to any one of the above-mentioned embodiments when executed by a processor.

[0162] The specific embodiment of the computer storage medium of the present application is basically the same as the above-mentioned mining method of customer behavior, and will not be repeated here.

[0163] It should be noted that in this paper, the term "including", "containing" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or system. Without more limitations, the element defined by the sentence "including a" does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0164] The above-mentioned embodiment number of the present application is only for description, not representing the pros and cons of the embodiment.

[0165] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, which is stored in a storage medium (such as ROM / RAM, magnetic disc, optical disc) as mentioned above, and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server or network device, etc.) execute the method described in each embodiment of the present application.

[0166] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent flow transformation made by using the content of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method of mining client behavior, characterized by, The mining method of the customer behavior comprises: obtaining historical information of lending and borrowing products of each of the customers to be mined; aggregating the historical information of the lending and borrowing products with preset crowd characteristic data to divide each of the customers to be mined into a plurality of customer groups; mining the lending and borrowing behaviors of the plurality of customer groups; the step of mining the lending and borrowing behaviors of the plurality of customer groups comprises: in a target customer group of the plurality of customer groups, determining current lending and borrowing product transactions of a target mining customer to generate a current transaction data set; learning the current transaction data set according to a preset association rule to generate a frequent lending and borrowing product set; based on the frequent lending and borrowing product set, mining the interdependence between the lending and borrowing products of the target mining customer; the step of mining the interdependence between the lending and borrowing products of the target mining customer based on the frequent lending and borrowing product set comprises: generating association rules based on the frequent lending and borrowing product set; determining the interdependence between the lending and borrowing products of the target mining customer according to target rules meeting a preset confidence condition from all the generated association rules.

2. The client behavior mining method of claim 1, wherein, the step of mining the lending and borrowing behaviors of the plurality of customer groups comprises: in a target customer group of the plurality of customer groups, determining lending and borrowing sequence data of a target mining customer; determining a latest frequent sequence pattern according to the lending and borrowing sequence data; based on the latest frequent sequence pattern, mining the sequence relationship of the lending and borrowing behaviors of the target mining customer.

3. The client behavior mining method of claim 2, wherein, the step of determining the lending and borrowing sequence data of the target mining customer comprises: determining historical basic data of the target mining customer; generating the lending and borrowing sequence data of the target mining customer according to the historical basic data.

4. The client behavior mining method of claim 1, wherein, the step of aggregating the historical information of the lending and borrowing products with preset crowd characteristic data to divide each of the customers to be mined into a plurality of customer groups comprises: generating a lending and borrowing history matrix according to the historical information of the lending and borrowing products; performing Gaussian clustering processing on a customer historical lending and borrowing behavior vector obtained by vectorizing the lending and borrowing history matrix and the preset crowd characteristic data to divide each of the customers to be mined into a plurality of customer groups.

5. A client behavior mining apparatus characterized by comprising: The mining device of the customer behavior comprises: an information obtaining module configured to obtain historical information of lending and borrowing products of each of the customers to be mined; a customer grouping module configured to aggregate the historical information of the lending and borrowing products with preset crowd characteristic data to divide each of the customers to be mined into a plurality of customer groups; a behavior mining module configured to mine the lending and borrowing behaviors of the plurality of customer groups; the behavior mining module comprises: a product dimension mining unit configured to, in a target customer group of the plurality of customer groups, determine current lending and borrowing product transactions of a target mining customer to generate a current transaction data set, learn the current transaction data set according to a preset association rule to generate a frequent lending and borrowing product set, and mine the interdependence between the lending and borrowing products of the target mining customer based on the frequent lending and borrowing product set; the product dimension mining unit comprises: a rule generation subunit configured to generate association rules based on the frequent lending and borrowing product set; The mining subunit is configured to determine the interdependence between the lending and borrowing products of the target mining client according to target rules that meet a preset confidence condition among all generated association rules.

6. The client behavior mining apparatus of claim 5, wherein The behavior mining module further includes: The time dimension mining unit is configured to determine the lending and borrowing sequence data of the target mining client in a target customer group among the plurality of customer groups, determine a latest frequent sequence pattern according to the lending and borrowing sequence data, and mine the lending and borrowing behavior sequence of the target mining client based on the latest frequent sequence pattern. The time dimension mining unit is further configured to determine historical basic data of the target mining client, and generate the lending and borrowing sequence data of the target mining client according to the historical basic data. The customer grouping module includes: A matrix generation unit is configured to generate a lending and borrowing history matrix according to the lending and borrowing product historical information. A clustering processing unit is configured to perform Gaussian clustering processing on a customer historical lending and borrowing behavior vector obtained by vectorizing the lending and borrowing history matrix and preset customer group feature data, so as to divide the plurality of to-be-mined customers into a plurality of customer groups.

7. A terminal device, characterized by, The terminal device includes a memory, a processor, and a customer behavior mining program stored in the memory and executable on the processor. When the customer behavior mining program is executed by the processor, the steps of the customer behavior mining method according to any one of claims 1 to 4 are implemented.

8. A computer storage medium, characterized in that The computer storage medium stores a customer behavior mining program. When the customer behavior mining program is executed by the processor, the steps of the customer behavior mining method according to any one of claims 1 to 4 are implemented.

Images