A financial product purchase prediction method, device and electronic equipment

Abstract

The present invention provides a financial product purchase prediction method, device and electronic device, which can be used in the financial field or other technical fields. The method comprises: inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for the target financial product and a pre-selected score for each customer in the customer group with purchase expectations; determining the attribute score of each customer according to the attribute information of each customer in the customer group with purchase expectations; determining the customer's purchase expectation value for the target financial product according to the pre-selected score and the attribute score of each customer. The financial product purchase prediction method, device and electronic device provided by the present invention can obtain the customer's purchase expectation value for the target financial product, thereby realizing the precise positioning of the customer group of the target financial product and marketing precisely matched financial products to specific customers.

Description

Technical Field

[0001] The present invention relates to the financial field or other technical fields, and in particular to a financial product purchase prediction method, device and electronic equipment. Background Art

[0002] The sale of personal financial products is one of the core tasks of commercial banks and an important source of profit. Generally speaking, personal financial products mainly include personal deposits and loans, funds, wealth management, insurance, foreign exchange, precious metals and bank cards. In recent years, commercial banks have launched a fierce competition in the field of personal finance, using big data and artificial intelligence technologies to accelerate the process of business transformation, hoping to be the first to wake up and seize the commanding heights in the context of the rapidly changing business environment.

[0003] In the past, commercial banks waited for customers to come to the outlets to express their financial needs, and then sold specific products. However, it seems that this can no longer meet the needs of banks and customers. Banks must not only tap into the resources of existing customers, conduct in-depth marketing, and fully stimulate the purchasing potential of each customer. They must also tap into the assets of customers who have switched to other banks to maximize overall performance. However, faced with millions of individual customers, how can a professional marketing team of more than a thousand people market precisely matched financial products to specific customers under the premise of strictly controlling costs, so as to meet customer needs? This has become a pain point that contemporary commercial banks have to face. Summary of the invention

[0004] In view of the problems in the prior art, the embodiments of the present invention provide a financial product purchase prediction method, device and electronic device, which can at least partially solve the problems in the prior art.

[0005] On the one hand, the present invention proposes a method for predicting the purchase of a financial product, the method comprising: inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for a target financial product and a pre-selected score for each customer in the customer group with purchase expectations; determining an attribute score for each customer based on the attribute information of each customer in the customer group with purchase expectations; determining the customer's expected purchase value for the target financial product based on the pre-selected score and the attribute score of each customer.

[0006] Optionally, before inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model, the method further includes: using the characteristic information of each customer in the target customer group who has purchased the target financial product as a training label, inputting the characteristic information of each customer in the sample customer group into the original model, and training to obtain the customer group pre-selection model, wherein the customers in the target customer group and the customers in the sample customer group are screened from the original customer group.

[0007] Optionally, the attributes of the customer include at least one of the following: inherent attributes, financial attributes, and purchasing behavior attributes; determining the attribute score of each customer in the customer group with purchasing expectations based on the attribute information of each customer includes: obtaining the inherent attribute score of each customer in the customer group with purchasing expectations based on the inherent attributes of each customer; and / or obtaining the financial attribute score of each customer in the customer group with purchasing expectations based on the financial attributes of each customer in the customer group with purchasing expectations; and / or obtaining the purchasing behavior attribute score of each customer in the customer group with purchasing expectations based on the purchasing behavior attributes of each customer.

[0008] Optionally, the inherent attributes of the customer include at least one of the following: name, gender, age, education level, height, weight, ID number, mobile phone number, and occupation; obtaining the inherent attribute score of each customer based on the inherent attributes of each customer in the customer group with purchase expectations includes: calculating the score of each attribute value of each inherent attribute based on the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product; for each customer in the customer group with purchase expectations, taking the product of the scores of the attribute values ​​of each inherent attribute of the customer as the inherent attribute score of the customer.

[0009] Optionally, the financial attributes of the customer include benefit-based attributes and cost-based attributes; and obtaining a financial attribute score of each customer in the customer group with purchase expectations according to the financial attribute of each customer includes: calculating the financial attribute score of each customer according to the following financial attribute score calculation formula:

[0010] in,

[0011] Fa represents the financial attribute score;

[0012] S i represents the value of the i-th benefit attribute;

[0013] D i Represents the i-th cost attribute value;

[0014] n 1 Represents the total number of benefit-type attributes;

[0015] n 2 Indicates the total number of cost attributes.

[0016] Optionally, the customer's purchase behavior attribute includes at least one of the following: the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product; the acquiring the purchase behavior attribute score of each customer in the customer group with purchase expectations according to the purchase behavior attribute includes: calculating the purchase behavior attribute score of each customer according to the following purchase behavior attribute scoring formula:

[0017] in,

[0018] Be represents the purchase behavior attribute score;

[0019] L i represents the attribute value of the i-th purchase behavior;

[0020] n represents the total number of purchase behavior attributes.

[0021] Optionally, determining the customer's expected purchase value for the target financial product based on the attribute score and pre-selected score of each customer includes: determining the customer's expected purchase value for the target financial product based on the pre-selected score, inherent attribute score and / or financial attribute score and / or purchase behavior attribute score of each customer.

[0022] Optionally, after determining the customer's expected purchase value of the target financial product according to the pre-selected score and the attribute score of each customer, the method further includes:

[0023] Determining a recommended customer group for the target financial product based on each customer's expected purchase value of the target financial product;

[0024] The target financial product is marketed to the customers in the recommended customer group.

[0025] On the other hand, the present invention provides a financial product purchase prediction device, comprising: an acquisition module, used to input the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model, to obtain a customer group with purchase expectations for the target financial product and a pre-selected score for each customer in the customer group with purchase expectations; a first determination module, used to determine the attribute score of each customer in the customer group with purchase expectations based on the attribute information of each customer in the customer group with purchase expectations; and a second determination module, used to determine the customer's expected purchase value for the target financial product based on the pre-selected score and the attribute score of each customer.

[0026] Optionally, the device also includes: a model training module, which is used to use the characteristic information of each customer in the target customer group who has purchased the target financial product as a training label, input the characteristic information of each customer in the sample customer group into the original model, and train to obtain the customer group pre-selection model, wherein the customers in the target customer group and the customers in the sample customer group are screened from the original customer group.

[0027] Optionally, the attributes of the customer include at least one of the following: inherent attributes, financial attributes, and purchasing behavior attributes; the first determination module includes: a first acquisition unit, used to acquire the inherent attribute score of each customer in the customer group with purchase expectations based on the inherent attributes of each customer; and / or a second acquisition unit, used to acquire the financial attribute score of each customer in the customer group with purchase expectations based on the financial attributes of each customer in the customer group with purchase expectations; and / or a third acquisition unit, used to acquire the purchasing behavior attribute score of each customer in the customer group with purchase expectations based on the purchasing behavior attributes of each customer.

[0028] Optionally, the inherent attributes of the customer include at least one of the following: name, gender, age, education level, height, weight, ID number, mobile phone number, and occupation; the first acquisition unit is specifically used to: calculate the score of each attribute value of each inherent attribute according to the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product; for each customer in the customer group with purchase expectations, the product of the scores of the attribute values ​​of each inherent attribute of the customer is used as the inherent attribute score of the customer.

[0029] Optionally, the financial attributes of the customer include benefit-based attributes and cost-based attributes; the second acquisition unit is specifically configured to calculate the financial attribute score of each customer according to the following financial attribute score calculation formula:

[0030] in,

[0031] Fa represents the financial attribute score;

[0032] S i represents the value of the i-th benefit attribute;

[0033] D i Represents the i-th cost attribute value;

[0034] n 1 Represents the total number of benefit-type attributes;

[0035] n 2 Indicates the total number of cost attributes.

[0036] Optionally, the customer's purchase behavior attribute includes at least one of the following: the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product; the third acquisition unit is specifically used to calculate the purchase behavior attribute score of each customer according to the following purchase behavior attribute scoring formula:

[0037] in,

[0038] Be represents the purchase behavior attribute score;

[0039] L i represents the attribute value of the i-th purchase behavior;

[0040] n represents the total number of purchase behavior attributes.

[0041] Optionally, the second determination module is specifically used to determine the customer's expected purchase value of the target financial product based on the pre-selected score, inherent attribute score and / or financial attribute score and / or purchase behavior attribute score of each customer.

[0042] Optionally, the device further includes: a third determination module, used to determine a recommended customer group for the target financial product based on the expected purchase value of the target financial product by each customer; and a marketing module, used to market the target financial product to customers in the recommended customer group.

[0043] On the other hand, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the financial product purchase prediction method described in any of the above embodiments when executing the program.

[0044] In yet another aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the financial product purchase prediction method described in any of the above embodiments.

[0045] The financial product purchase prediction method, device and electronic device provided in the embodiments of the present invention can preliminarily screen out customer groups with certain purchase expectations for target financial products through machine learning methods, and then use statistical learning methods to describe the correlation between customer attribute information and customer purchase events of target financial products, thereby making up for the shortcomings of machine learning in having strict requirements on data quality. Finally, the pre-selected score of each customer obtained by machine learning and the attribute score of each customer obtained by statistical learning are combined to obtain the purchase expectation value of each customer for the target financial product, thereby achieving accurate positioning of the customer group of the target financial product and marketing of precisely matched financial products to specific customers.

[0046] From the perspective of banking business, each type of personal financial product can accurately target the customer base, and the goal of reducing costs and increasing efficiency can be achieved. In the process, customers' preferences can also be learned and mastered, providing a solid foundation for further improving customer maintenance and management effects. From the customer's perspective, each customer can get a list of recommended products that best suits them, meeting the different purchasing needs of the customer base, thereby increasing customer satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. 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 creative work. In the drawings:

[0048] Figure 1 It is a flowchart of a financial product purchase prediction method provided by an embodiment of the present invention.

[0049] Figure 2 It is a flowchart of a financial product purchase prediction method provided by another embodiment of the present invention.

[0050] Figure 3 It is a schematic diagram of the structure of a financial product purchase prediction device provided by an embodiment of the present invention.

[0051] Figure 4 It is a structural diagram of a financial product purchase prediction device provided by another embodiment of the present invention.

[0052] Figure 5 It is a structural diagram of a first determination module of a financial product purchase prediction device provided by another embodiment of the present invention.

[0053] Figure 6 It is a structural diagram of a financial product purchase prediction device provided by yet another embodiment of the present invention.

[0054] Figure 7 It is a schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention. It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other arbitrarily.

[0056] The execution subject of the financial product purchase prediction method provided by the embodiment of the present invention includes but is not limited to a computer.

[0057] Figure 1 FIG. 1 is a flow chart of a method for predicting purchase of financial products provided by an embodiment of the present invention. Figure 1 As shown, the financial product purchase prediction method provided by the embodiment of the present invention includes:

[0058] S101, inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for the target financial product and a pre-selection score for each customer in the customer group with purchase expectations;

[0059] In this step, the original customer group may be an unscreened customer group of the banking system within the target time period, such as a customer group within the past 10 years; the customer feature information may include the customer's inherent attribute information and financial attribute information, such as the customer's inherent attributes may include gender (Se), age (Ag), education (Ed) and occupation (Oc), etc., and the financial attributes may include banking service star rating (St), daily average financial assets (Da) and personal loan balance at a certain point in time (Lb), etc. The target financial products may be personal deposits and loans, funds, wealth management, insurance, foreign exchange, precious metals and bank cards, etc.

[0060] The pre-trained customer group pre-selection model is used to calculate the pre-selection score of each customer in the original customer group and classify the customers in the original customer group, so as to preliminarily screen out the customers who have the purchase expectation for the target financial product and obtain the pre-selection score of the customers who have the purchase expectation. Each of the customers who have the purchase expectation for the target financial product constitutes the customer group with the purchase expectation (customer group to be studied). The customer group with the purchase expectation may include both customers who have purchased the target financial product and customers who have never purchased the target financial product.

[0061] The customer group pre-selection model can be trained using a machine learning algorithm. The advantage of a machine learning algorithm is the processing, learning and classification of massive data. However, its actual effect is subject to many factors such as the richness of customer feature dimensions, the size of positive sample coverage, data quality and the experience of modeling engineers. This step uses a machine learning method to pre-learn massive customer data and screen out customer groups (customer groups to be studied) that have certain purchase expectations for the target financial products. Subsequently, a statistical learning method can be used to re-optimize the pre-selection results.

[0062] S102, determining an attribute score for each customer in the customer group with purchase expectations according to the attribute information of each customer;

[0063] In this step, the customer's attribute information may include inherent attributes, financial attributes, and purchase behavior attributes. Since there may be a correlation between the customer's purchase behavior of the target financial product and the above-mentioned attribute information of the customer, the attribute score of each customer in the customer group with purchase expectations (the customer group to be studied) can be determined based on statistical methods combined with expert experience, which can effectively supplement the machine learning method. This step is essentially a quantitative description of the correlation between the customer attribute information and the customer's purchase of the target financial product.

[0064] S103: Determine the customer's expected purchase value for the target financial product based on the pre-selected score and the attribute score of each customer.

[0065] In this step, the pre-selected scores obtained by machine learning methods and the attribute scores obtained by statistical methods complement each other. Machine learning can process massive amounts of data and fully examine the correlation between customer characteristic information and the purchase of target financial products, but its effectiveness depends to a large extent on data quality and the experience of model builders. It is a black-box learning model. Statistical learning adopts a white-box perspective and focuses on a limited number of attribute information, downplays the strict requirements for the quality of all data, and makes use of expert experience to make up for the shortcomings of machine learning, ultimately obtaining reliable purchase expectations.

[0066] The purchase expectation value obtained according to the pre-selection score and the attribute score may have a value range between 0 and 1, and the purchase expectation value of each customer may represent the probability of the customer purchasing the target financial product. Specifically, the pre-selection score and the attribute score may have different corresponding weights, and the purchase expectation value may be obtained by weighted calculation based on the pre-selection score and the attribute score and the weight values ​​of the two scores.

[0067] The financial product purchase prediction method provided by the embodiment of the present invention can preliminarily screen out customer groups with certain purchase expectations for the target financial product through a machine learning method, and then use a statistical learning method to describe the correlation between customer attribute information and the event of the customer purchasing the target financial product, thereby making up for the shortcoming of machine learning in having strict requirements on data quality. Finally, the pre-selected score of each customer obtained by machine learning and the attribute score of each customer obtained by statistical learning are combined to obtain the purchase expectation value of each customer for the target financial product, thereby achieving accurate positioning of the customer group of the target financial product and marketing of precisely matched financial products to specific customers.

[0068] From the perspective of banking business, each type of personal financial product can accurately target the customer base, and the goal of reducing costs and increasing efficiency can be achieved. In the process, customers' preferences can also be learned and mastered, providing a solid foundation for further improving customer maintenance and management effects. From the customer's perspective, each customer can get a list of recommended products that best suits them, meeting the different purchasing needs of the customer base, thereby increasing customer satisfaction and loyalty.

[0069] Optionally, in the above embodiment, before inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model, the method further includes: using the characteristic information of each customer in the target customer group who has purchased the target financial product as a training label, inputting the characteristic information of each customer in the sample customer group into the original model, and training to obtain the customer group pre-selection model, wherein the customers in the target customer group and the customers in the sample customer group are screened from the original customer group.

[0070] In this embodiment, every customer in the target customer group has purchased the target financial product. Optionally, a customer group that has purchased the target financial product within the past y (y is a natural number) years may be designated as the target customer group.

[0071] The training process of the customer group pre-selection model can be as follows:

[0072] Step 1: Select positive samples for machine learning (target customer groups). You can specify a subset of customers who have purchased target financial products (e.g., stock funds) in the past y years as positive samples for machine learning. The positive sample set can be filtered by the total amount of the target financial products purchased by the customers, for example, customers who have purchased such financial products for a total amount greater than 100,000 in the past y years;

[0073] Step 2: Determine feature information and feature engineering. In the field of machine learning, it is generally believed that the richer the number of features, the more accurate the prediction of customer purchasing behavior. When selecting features, you can focus on the following two types of features: customer inherent attributes and customer financial attributes. Among them, customer inherent attributes refer to attributes that are not related to finance, including at least gender (Se), age (Ag), education (Ed) and occupation (Oc). Customer financial attributes include at least bank service star rating (St), average daily financial assets (Da) and personal loan balance at a certain point in time (Lb).

[0074] Extract customer data according to the selected feature dimensions, perform feature engineering, quantify all data (for example, convert gender into 1 or 2), supplement missing data (for example, randomly supplement), and correct erroneous data (for example, if gender is unknown, it can be randomly assigned 1, 2, or 3).

[0075] Step 3: Conduct model training. After determining the data dimension, data content, positive samples, training set, and test set, you can try to use a variety of different machine learning algorithms, and finally determine a classic algorithm with the highest "area under the receiver operating characteristic curve" (AUC) score (for example: adaptive boosting tree algorithm (AdaBoost) or gradient boosting tree algorithm (GBDT)). The results of the algorithm execution will classify the customers in the training set and test set into two categories, that is, customers who are likely to purchase this type of personal financial product or those who are unlikely to purchase it.

[0076] It should be pointed out here that the advantage of machine learning algorithms lies in the processing, learning and classification of massive data. However, its actual effect is subject to many factors such as the richness of customer feature dimensions, the coverage of positive samples (target customer groups), data quality and the experience of modeling engineers. The present invention uses machine learning methods to pre-learn massive customer data and screen out customer groups with certain purchase expectations (hereinafter referred to as "customer groups to be studied"). Subsequently, statistical learning methods are used to re-optimize the pre-selected results.

[0077] Optionally, in any of the above embodiments, the customer attributes include at least one of the following: inherent attributes, financial attributes, and purchase behavior attributes; and determining the attribute score of each customer in the customer group with purchase expectations according to the attribute information of each customer includes:

[0078] According to the inherent attributes of each customer in the customer group with purchase expectations, obtaining an inherent attribute score of each customer; and / or

[0079] According to the financial attribute of each customer in the customer group with purchase expectations, obtaining a financial attribute score for each customer; and / or

[0080] According to the purchase behavior attribute of each customer in the customer group with purchase expectations, a purchase behavior attribute score of each customer is obtained.

[0081] In this embodiment, the inherent attributes of the customer refer to some attributes of the customer itself, which are irrelevant to the business the customer has handled in the banking system, such as name, gender, age, education, height, weight, ID number, mobile phone number, occupation, etc. The financial attributes of the customer are related to the financial business the customer has handled in the banking system, such as bank service star rating (St), daily average financial asset amount (Da) and personal loan balance at a certain point in time (Lb); the purchase behavior attributes of the customer are related to the customer's purchase behavior of the target financial product, such as the latest purchase time (Lt), the number of months of holding such products (Tm), the total purchase amount (Ta), the total purchase frequency (Tf, number of times), etc.

[0082] According to the definitions of inherent attributes, financial attributes and purchase behavior attributes, each type of attribute has different effects on the customer's purchase of target financial products. For example, for the inherent attributes of the customer, the customer's gender, age, education and occupation may have a greater impact on the customer's purchase of the target financial product; and for the customer's financial attributes, the customer's banking service star directly reflects the customer's comprehensive asset situation in the commercial bank. The higher the star rating, the more assets, and vice versa. In addition, this indicator also represents the customer's contribution to the commercial bank. The higher the star rating, the more bank financial products the customer holds, and vice versa. From a statistical perspective, the higher the customer's contribution star rating, the greater the possibility of purchasing personal financial products, and vice versa. St is a benefit-type indicator for the customer's expected purchase evaluation, and Da is also a benefit-type indicator; Lb reflects the total amount of loans that the customer has not yet repaid at a certain moment. The higher its value, the more debt the customer has, the less disposable cash, and the less likely he is to purchase various financial products. On the contrary, the larger the value, the smaller the possibility of purchasing various financial products. It can be seen that Lb is a cost-type indicator. As for the customer's purchase behavior attributes, the larger the Lt (format is yyyymmdd), the closer the customer's last purchase time is to the present time. Considering that the enthusiasm for purchase has not dissipated, the possibility of repurchase in the near future is relatively high, so Lt is a benefit indicator. The larger the Tm, the higher the customer's love and trust in the purchased financial products, and vice versa. The lower the Tm, the same in nature. Ta and Tf are obviously benefit indicators. Therefore, according to the customer's inherent attributes, financial attributes, and purchase behavior attributes, the correlation between each type of attribute and the customer's purchase of the target financial product can be analyzed by statistical learning methods according to the attribute category.

[0083] Optionally, in the above embodiment, the inherent attributes of the customer may include at least one of the following: name, gender, age, education, height, weight, ID number, mobile phone number, occupation; and obtaining the inherent attribute score of each customer in the customer group with purchase expectations according to the inherent attributes of each customer may include:

[0084] Calculate the score of each attribute value of each inherent attribute according to the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product;

[0085] For each customer in the customer group with purchase expectations, the product of the scores of the attribute values ​​of each inherent attribute of the customer is taken as the inherent attribute score of the customer.

[0086] In this embodiment, we first define the inherent attributes of customers, that is, attributes that are not related to financial characteristics, such as name, gender, age, education, height, weight, ID number, mobile phone number, occupation, etc. In statistical learning, experts need to select multiple inherent attributes that are highly correlated with the purchase of such financial products based on their own experience, that is, core inherent attributes. The core inherent attributes selected in the present invention are gender, age, education and occupation.

[0087] Taking the core inherent attributes of gender, age, education and occupation as an example, the calculation of the score of each attribute value of each inherent attribute according to the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product includes:

[0088] Step 1: First, quantify these four core intrinsic attributes.

[0089] Gender = {Male, Female, Unknown};

[0090] Education = {less than college, college, bachelor's degree, master's degree, doctorate, postdoctoral fellow};

[0091] Occupation = {IT, Finance, Education, Culture and Arts, Military, Medical, Other State-owned, Other Private, Foreign-owned};

[0092] For continuous intrinsic attributes, they should be converted into discrete ones through interval partitioning:

[0093] Age = {minor (<18 years old), young ([18-45] years old), middle-aged ([46-69] years old), elderly (>69 years old)}.

[0094] After simple quantification, the four customer-specific attribute sets can be expressed as follows:

[0095] Gender Se = {1, 2, 3}; a certain attribute value can be represented by Sej, where j represents the jth attribute of Se, and the following is similar.

[0096] Education levelEd={1,2,3,4,5,6};

[0097] Occupation Oc = {1,2,3,4,5,6,7,8,9};

[0098] Age Ag = {1, 2, 3, 4};

[0099] Step 2: Intrinsic attribute value score assessment.

[0100] According to the machine learning positive samples (target customer groups) selected in the above embodiment, mapping is performed according to different inherent attributes and their values. The specific method is as follows: for the jth attribute value of attribute i, if there are x customers in the positive sample set corresponding to attribute value j, then these x customers are said to have completed the mapping with j, and the mapping score of attribute value j (P ij )for:

[0101]

[0102] Where A is the positive sample set, N(A) is the total number of customers in set A, and P ij is the score of the jth attribute value of the inherent attribute i. When examining the inherent attribute evaluation score (inherent attribute score) of a customer to be studied, if the customer's attribute matches ij, the corresponding score P can be obtained. ij To make it easier to understand, let's take an example. For example, in set A, there are 28% males, 70% females, and 2% unknown customers. Then the score of "male" is 0.28, the score of "female" is 0.70, and the score of "unknown" is 0.02. That is to say, after the mapping is completed, each attribute value of the inherent attribute has a corresponding score. This score shows the situation of customers with this inherent attribute value purchasing the target financial product.

[0103] From the above example, we can see that the number of female customers in set A is 2.5 times that of male customers. In other words, if the customer in the customer group to be studied is female, she will get a score of 0.70 on the Se intrinsic attribute. This score will be used in subsequent calculations. Similarly, we can get the score of a customer on Ag, Ed and Oc. This step is essentially a quantitative description of the correlation between the intrinsic attribute value and the event of the customer purchasing the target financial product.

[0104] After obtaining the scores of the attribute values ​​of each inherent attribute according to the method in the above embodiment, for each customer in the customer group with purchase expectations, taking the product of the scores of the attribute values ​​of each inherent attribute of the customer as the inherent attribute score of the customer includes:

[0105] For each customer in the customer group with purchase expectations (customer group to be studied), one of the four inherent attributes is corresponding, and the score of the customer on the four inherent attributes is obtained and recorded as S 1 , S 2 , S 3 and S 4 Then the comprehensive score of this customer on the inherent attributes (inherent attribute score) can be defined by the following formula:

[0106]

[0107] In represents the inherent attribute score of a customer, S i represents the score of a customer on the intrinsic attribute i. δ is the value space adjustment constant, S i The value space of is [0,1] (expressed as a percentage), n is the number of the inherent attributes of the customer, in this embodiment n = 4, then the value obtained after multiplication may be very small, and an adjustment constant is required, for example, δ can be preset to 10 4 .

[0108] Optionally, when the financial attributes of the customer include benefit-based attributes and cost-based attributes, obtaining the financial attribute score of each customer in the customer group with purchase expectations according to the financial attribute includes: calculating the financial attribute score of each customer according to the following financial attribute score calculation formula:

[0109] in,

[0110] Fa represents the financial attribute score;

[0111] S i represents the value of the i-th benefit attribute;

[0112] D i Represents the i-th cost attribute value;

[0113] n 1 Represents the total number of benefit-type attributes;

[0114] n 2 Indicates the total number of cost attributes.

[0115] In this embodiment, the financial attributes of the customer may include the bank service star rating (St), the average daily financial assets (Da), and the personal loan balance at a certain point in time (Lb). First, analyze the St indicator. The bank service star rating directly reflects the comprehensive asset situation of the customer in the commercial bank. The higher the star rating, the more assets, and vice versa. In addition, this indicator also represents the customer's contribution to the commercial bank. The higher the star rating, the more bank financial products the customer holds, and vice versa. From a statistical perspective, the higher the customer's contribution star rating, the greater the possibility of purchasing personal financial products, and vice versa. St is a benefit attribute of the customer's expected purchase evaluation, and Da is also a benefit attribute; Lb reflects the total amount of loans that the customer has not yet repaid at a certain moment. The higher its value, the more debt the customer has, the less disposable cash, and the less likely he is to purchase various financial products, and vice versa. It can be seen that Lb is a cost attribute.

[0116] Therefore, the customer's financial attribute score can be obtained by multiplying the product of the benefit attribute values ​​in the customer's financial attributes by the cost attribute values ​​(see the above financial attribute score calculation formula).

[0117] The following takes the customer's benefit attributes including banking service star rating (St) and average daily financial asset amount (Da) and the cost attributes including the personal loan balance at a certain point in time (Lb) as an example to describe in detail the process of obtaining the financial attribute score of each customer in the customer group with purchase expectations based on the financial attributes of each customer:

[0118] Step 1: Extract the financial attribute data of each customer in the customer group with purchase expectations, including bank service star rating (St), daily average financial assets (Da), and personal loan balance at a certain point in time (Lb).

[0119] Step 2: Customer financial attribute evaluation method.

[0120] Below we give the calculation method of the financial attribute assessment value (Fa) of a customer:

[0121]

[0122] As can be seen from the above formula, Fa is obtained by multiplying the two benefit attribute values ​​(St and Da) in the financial attribute by the cost attribute value (Lb). Since the values ​​of Da and Lb may be large or small between different customers, it is necessary to set an adjustment constant θ to adjust the value of Fa. The value of θ should be flexibly set according to the calculation results before adjustment. The adjustment constant φ is a preset small positive number (for example, 1), which is mainly used to prevent the calculation error of the financial attribute evaluation value caused by a customer's Lb being 0.

[0123] In the following steps, we will optimize the calculation of Fa, compress its value space, and improve the overall operating efficiency.

[0124] Step 3: Optimize the value space compression.

[0125] According to the general experience of commercial banks, the value space of St is a positive integer from 1 to 7, while the value spaces of Da and Lb are both positive real numbers in [0, +∞). From the actual data, the difference between these two values ​​​​is very large between different customers. The three attribute values ​​​​used in evaluating Fa are obviously not in the same dimension, which is not conducive to understanding and practical application. Therefore, the present invention designs a compression method for the value space of Fa to control its value within a limited range.

[0126] First, Da and Lb are converted into discrete variables, such as {very low (<10,000), low ([1,100,000), average ([10,300,000), high [300,1000,000]), very high (>1000,000)}. Then they are quantized into a set of numbers {1,2,3,4,5}. The participation of these values ​​in the calculation can effectively compress the value space of Fa and increase the interpretability and practicality of the calculation results.

[0127] The calculation method of Fa is optimized as follows:

[0128]

[0129] Where Da′ and Lb′ are the compressed average daily financial assets and loan balances of customers at that time. It can be seen that the value space of Fa after compression is [1,35], and two adjustment constants θ and φ are omitted, which effectively reduces the complexity of calculation.

[0130] This compression algorithm may cause certain data information to disappear. For example, after compression, the device believes that the evaluation effectiveness of two customers with daily average financial assets of 20 million and 100 million is the same on the Da indicator. However, in actual marketing scenarios, we found that the evaluation calculation of customer financial attributes only needs to maintain this granularity to effectively distinguish customer purchase expectations. Too high precision may have a negative impact on interpretability and ease of use.

[0131] Optionally, in any of the above embodiments, the customer's purchase behavior attribute includes at least one of the following: the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product; and obtaining the purchase behavior attribute score of each customer in the customer group with purchase expectations according to the purchase behavior attribute of each customer includes:

[0132] The purchase behavior attribute score of each customer is calculated according to the following purchase behavior attribute score formula:

[0133] in,

[0134] Be represents the purchase behavior attribute score;

[0135] L i represents the attribute value of the i-th purchase behavior;

[0136] n represents the total number of purchase behavior attributes.

[0137] In this embodiment, for customers in the customer group with purchase expectations, some customers may have purchased the target financial product, while other customers have no purchase behavior. For these customers without purchase behavior, their purchase behavior attribute score is 0.

[0138] The latest purchase time Lt of the target financial product (in the format of yyyymmdd), the larger its value is, the closer the customer's last purchase time is to the present time. Considering that the enthusiasm for purchase has not yet dissipated, the possibility of repurchase in the near future is relatively high, so Lt is a benefit indicator; the larger the holding time Tm of the target financial product is, the higher the customer's love and trust in the purchased financial product is, and vice versa, Tm is also a benefit indicator; Ta and Tf are obviously both benefit indicators. Therefore, the product of the customer's purchase behavior attribute values ​​can be used as the customer's purchase behavior attribute score (see the above purchase behavior attribute score formula).

[0139] When the customer's purchase behavior attributes include the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product, a preferred mathematical definition of the customer purchase behavior attribute score (Be) is given below:

[0140] Be=ε(Lt×Tm×Ta×Tf), ε=1 / 20million;

[0141] Where ε is an adjustment constant, which can be preset to 20 million, mainly to regulate the value of Lt in [0,10). If one of the four attributes is 0, Be is 0, which means that the customer has not purchased the target financial product. The value range of Be is [0,+∞). Since the real data verifies that the values ​​of Ta and Tf are rarely too large, only the ε constant is used for adjustment, and there is no need to use the value space compression method.

[0142] Optionally, when the attributes of the customer include inherent attributes and / or financial attributes and / or purchasing behavior attributes, determining the customer's expected purchase value for the target financial product based on the attribute score and pre-selected score of each customer includes: determining the customer's expected purchase value for the target financial product based on the pre-selected score, inherent attribute score and / or financial attribute score and / or purchasing behavior attribute score of each customer.

[0143] In this embodiment, the pre-selected score of each customer is obtained by machine learning, while the inherent attribute score, financial attribute score and purchase behavior score of the customer are obtained based on statistical methods combined with expert experience. Therefore, by combining the pre-selected score obtained by machine learning and at least one inherent attribute score obtained by statistical methods, a quantitative and unique purchase expectation value is calculated for each customer, providing strong data and technical support for commercial banks to expand their customer base, expand their business, save costs and increase revenue.

[0144] Optionally, when the attributes of the customer include inherent attributes, financial attributes, and purchase behavior attributes, determining the customer's expected purchase value of the target financial product according to the attribute score and pre-selected score of each customer may specifically include the following steps:

[0145] Step 1: Standardization and normalization of customer pre-selection scores, inherent attributes, financial attributes, and purchase behavior attributes.

[0146] First, these four sets of data need to be standardized. Standardization means unifying the properties of these attributes (cost type, benefit type, etc.). We use Ml to represent the pre-selection score of machine learning. The larger the value, the greater the possibility that the customer will purchase the target financial product in the future, and vice versa. It can be seen that Ml is a benefit indicator. The larger the value of In, the higher the similarity between the customer's age, gender and occupation and the inherent attributes of customers who have purchased the target financial product, and vice versa. In, Fa and Be are also benefit indicators. The four indicators involved in the final decision evaluation are all benefit-type, and no additional standardization is required.

[0147] Then, the formats of the four sets of data are processed as necessary, so that the value ranges of Ml, In, Fa and Be are converted to (0,1], that is, normalization. (Ml has been normalized during the execution of the machine learning algorithm)

[0148]

[0149] where x i represents the i-th attribute value of one of the three customer attributes, n is the number of customers in the customer group with purchase expectations, x i ′ represents x i The value after normalization. η is a small positive number that can be preset to ensure that both the denominator and the numerator of the formula are not 0. At the same time, the value space of the four sets of data after normalization is (0,1].

[0150] Step 2: Integration of indicators.

[0151] This embodiment involves a total of 4 sub-evaluation indicators: Ml, In, Fa and Be, which can be divided into two categories: machine learning indicators and statistical learning indicators. The two types of indicators complement each other. Machine learning can process massive amounts of data and attribute systems and fully examine their correlations, but the effect depends largely on data quality and the experience of model builders. It is a black-box learning model. Statistical learning adopts a white-box perspective and only focuses on a few focused attributes, downplaying the strict requirements for the quality of all data and making up for the differences between modelers with the help of expert experience. Below we use the weighted average method to define the customer's expected purchase evaluation value E:

[0152]

[0153] wherein are the weights of each index respectively, and the weight values of each weight can be jointly determined by business experts and technical experts according to the different importance of the four indexes.

[0154] For example Figure 2 As shown, optionally, after determining the purchase expectation value of the target financial product for each customer according to the preselected score and attribute score of each customer, the method may further include:

[0155] S104. Determine the recommended customer group of the target financial product according to the purchase expectation value of each customer for the target financial product;

[0156] In this step, since the greater the purchase expectation value, the greater the probability that the customer will purchase the target financial product. Therefore, after obtaining the purchase expectation value of each customer in the customer group with purchase expectation, the customers with a purchase expectation value greater than the target threshold can be used as recommended customers to form the recommended customer group, or the customers in the customer group with purchase expectation can be sorted in descending order of the purchase expectation value, and the top N (N is a positive integer) customers can be used as recommended customers to form the recommended customer group, or the top a% (0 < a < 100) customers can be used as recommended customers to form the recommended customer group.

[0157] S105. Market the target financial product to the customers in the recommended customer group.

[0158] In this step, the target financial product can be recommended to the customer through the bank system APP installed by the customer on the client side, or the target financial product can be recommended to the customer by text message or phone call using the mobile phone number reserved by the customer, or the target financial product can be recommended to the customer by email using the email reserved by the customer.

[0159] Optionally, after determining the purchase expectation value of each customer for the target financial product according to the preselected score and attribute score of each customer, a list of the purchase expectation values of each customer in the customer group with purchase expectation can also be generated as needed. The front-line marketers of the commercial bank can simply conduct phone or on-site marketing in order after getting this list sorted in descending order of the purchase expectation value, eliminating the expenditure of other training or learning costs.

[0160] The financial product purchase prediction method provided by the embodiments of the present invention provides strong data and technical support for the commercial bank to expand customers and businesses, and increase revenue while reducing costs. The main effects and advantages of the present invention are as follows:

[0161] 1. Applicable to various personal financial product marketing scenarios. The scope of application of the present invention basically covers all personal financial product precision marketing scenarios of commercial banks. Of course, it is inevitable to tailor data according to the characteristics of the actual application scenarios. For example: for different marketing scenarios such as financial management, funds, insurance and credit cards, there must be differences in the original customer groups, positive sample sets and customer purchase behavior data. At this time, it is only necessary to extract different customer data according to the actual situation, without adjusting the algorithm system involved in the present invention.

[0162] 2. Use statistical learning methods to supplement the limitations of machine learning. In big data analysis scenarios, machine learning is often used as the main method to achieve precision marketing. However, the actual effect of this method is largely limited to the ideas and experience of the modeler. In addition to using machine learning methods, the present invention further designs and develops a set of statistical learning methods, fully considering the experience of business experts, and starting from the three categories of customer inherent attributes, customer financial attributes and customer purchasing behavior attributes, a total of eleven attributes that need to be focused on given by business experts, designing algorithms such as "inherent attribute evaluation" and "financial and purchasing behavior attribute evaluation", and finally integrating them with the pre-selected scores generated by the machine learning algorithm to calculate a quantitative and unique purchase expectation value for each customer with purchase expectations.

[0163] 3. The marketing strategy is simple and easy to implement. Another advantage of the present invention is that the algorithm is simple, easy to use, and highly understandable and operable. The method is designed to be scientific, accurate, and concise, and to avoid overly complex logic and calculations. The deployment and assembly process is very simple. The present invention generates the expected purchase value of the target financial product for customers with purchase expectations on demand. The front-line marketing personnel of commercial banks get this list of expected purchase values ​​arranged from large to small, and only need to conduct telephone or on-site marketing in order, eliminating the expenditure of other training or learning costs.

[0164] Figure 3 FIG. 1 is a schematic diagram of the structure of a financial product purchase prediction device provided by an embodiment of the present invention. Figure 3 As shown, the financial product purchase prediction device provided by the embodiment of the present invention includes: an acquisition module 21, which is used to input the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for the target financial product and a pre-selected score of each customer in the customer group with purchase expectations; a first determination module 22, which is used to determine the attribute score of each customer in the customer group with purchase expectations based on the attribute information of each customer; and a second determination module 23, which is used to determine the customer's purchase expectation value for the target financial product based on the pre-selected score and the attribute score of each customer.

[0165] The financial product purchase prediction device provided by the embodiment of the present invention can preliminarily screen out customer groups with certain purchase expectations for target financial products through machine learning methods, and then use statistical learning methods to describe the correlation between customer attribute information and customer purchase events of target financial products, thereby making up for the shortcomings of machine learning in having strict requirements on data quality. Finally, the pre-selected score of each customer obtained by machine learning and the attribute score of each customer obtained by statistical learning are combined to obtain the purchase expectation value of each customer for the target financial product, thereby achieving accurate positioning of the customer group of the target financial product and marketing of precisely matched financial products to specific customers.

[0166] From the perspective of banking business, each type of personal financial product can accurately target the customer base, and the goal of reducing costs and increasing efficiency can be achieved. In the process, customers' preferences can also be learned and mastered, providing a solid foundation for further improving customer maintenance and management effects. From the customer's perspective, each customer can get a list of recommended products that best suits them, meeting the different purchasing needs of the customer base, thereby increasing customer satisfaction and loyalty.

[0167] like Figure 4 As shown, optionally, the device also includes: a model training module 24, which is used to use the characteristic information of each customer in the target customer group who has purchased the target financial product as a training label, input the characteristic information of each customer in the sample customer group into the original model, and train to obtain the customer group pre-selection model, wherein the customers in the target customer group and the customers in the sample customer group are screened from the original customer group.

[0168] like Figure 5 As shown, optionally, the attributes of the customer include at least one of the following: inherent attributes, financial attributes, and purchase behavior attributes; the first determination module 22 includes: a first acquisition unit 221, used to obtain the inherent attribute score of each customer in the customer group with purchase expectations according to the inherent attributes of each customer; and / or a second acquisition unit 222, used to obtain the financial attribute score of each customer in the customer group with purchase expectations according to the financial attributes of each customer in the customer group with purchase expectations; and / or a third acquisition unit 223, used to obtain the purchase behavior attribute score of each customer in the customer group with purchase expectations according to the purchase behavior attributes of each customer.

[0169] Optionally, the inherent attributes of the customer include at least one of the following: name, gender, age, education level, height, weight, ID number, mobile phone number, and occupation; the first acquisition unit is specifically used to: calculate the score of each attribute value of each inherent attribute according to the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product; for each customer in the customer group with purchase expectations, the product of the scores of the attribute values ​​of each inherent attribute of the customer is used as the inherent attribute score of the customer.

[0170] Optionally, the financial attributes of the customer include benefit-based attributes and cost-based attributes; the second acquisition unit is specifically configured to calculate the financial attribute score of each customer according to the following financial attribute score calculation formula:

[0171] in,

[0172] Fa represents the financial attribute score;

[0173] S i represents the value of the i-th benefit attribute;

[0174] D i Represents the i-th cost attribute value;

[0175] n 1 Represents the total number of benefit-type attributes;

[0176] n 2 Indicates the total number of cost attributes.

[0177] Optionally, the customer's purchase behavior attribute includes at least one of the following: the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product; the third acquisition unit is specifically used to calculate the purchase behavior attribute score of each customer according to the following purchase behavior attribute scoring formula:

[0178] in,

[0179] Be represents the purchase behavior attribute score;

[0180] L i represents the attribute value of the i-th purchase behavior;

[0181] n represents the total number of purchase behavior attributes.

[0182] Optionally, the second determination module is specifically used to determine the customer's expected purchase value of the target financial product based on the pre-selected score, inherent attribute score and / or financial attribute score and / or purchase behavior attribute score of each customer.

[0183] like Figure 6 As shown, optionally, the device also includes: a third determination module 25, used to determine the recommended customer group of the target financial product according to the expected purchase value of each customer for the target financial product; a marketing module 26, used to market the target financial product to the customers in the recommended customer group.

[0184] The embodiments of the device provided by the present invention can be specifically used to execute the processing flow of the above-mentioned method embodiments. Its functions are not described in detail here, and reference can be made to the detailed description of the above-mentioned method embodiments.

[0185] It should be noted that the financial product purchase prediction method and device provided in the embodiments of the present invention can be used in the financial field, and can also be used in any technical field other than the financial field. The embodiments of the present invention do not limit the application field of the financial product purchase prediction method and device.

[0186] Figure 7 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 301, a communication interface 302, a memory 303 and a communication bus 304, wherein the processor 301, the communication interface 302 and the memory 303 communicate with each other through the communication bus 304. The processor 301 may call the logic instructions in the memory 303 to execute the following method: input the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for the target financial product and a pre-selection score of each customer in the customer group with purchase expectations; determine the attribute score of each customer according to the attribute information of each customer in the customer group with purchase expectations; determine the purchase expectation value of the customer for the target financial product according to the pre-selection score and the attribute score of each customer.

[0187] In addition, the logic instructions in the above-mentioned memory 303 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0188] The present embodiment discloses a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments, for example, including: inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for a target financial product and a pre-selected score for each customer in the customer group with purchase expectations; determining the attribute score of each customer in the customer group with purchase expectations based on the attribute information of each customer; and determining the customer's expected purchase value for the target financial product based on the pre-selected score and the attribute score of each customer.

[0189] The present embodiment provides a computer-readable storage medium, which stores a computer program. The computer program enables the computer to execute the methods provided by the above-mentioned method embodiments, for example, including: inputting the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for the target financial product and a pre-selected score for each customer in the customer group with purchase expectations; determining the attribute score of each customer in the customer group with purchase expectations based on the attribute information of each customer; determining the customer's expected purchase value for the target financial product based on the pre-selected score and the attribute score of each customer.

[0190] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention 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 code.

[0191] 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.

[0192] 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.

[0193] 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.

[0194] In the description of this specification, the description with reference to the terms "one embodiment", "a specific embodiment", "some embodiments", "for example", "example", "specific example", or "some examples" 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 present invention. In this specification, the schematic representation of the above terms does 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.

[0195] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting purchase of financial products, characterized in that: The method comprises: Input the characteristic information of each customer in the original customer group into the pre-trained customer group pre-selection model to obtain the customer group with purchase expectations for the target financial product and the pre-selection score of each customer in the customer group with purchase expectations; Determining an attribute score for each customer according to the attribute information of each customer in the customer group with purchase expectations; Determining the customer's expected purchase value of the target financial product based on the pre-selected score and the attribute score of each customer; The attributes of the customer include: inherent attributes, financial attributes, and purchase behavior attributes; Determining the attribute score of each customer in the customer group with purchase expectations according to the attribute information of each customer includes: obtaining the inherent attribute score of each customer in the customer group with purchase expectations according to the inherent attribute of each customer; obtaining the financial attribute score of each customer in the customer group with purchase expectations according to the financial attribute of each customer in the customer group with purchase expectations; and obtaining the purchase behavior attribute score of each customer in the customer group with purchase expectations according to the purchase behavior attribute of each customer; The financial attributes of the customer include benefit attributes and cost attributes. The benefit attributes include at least one of the following: bank service star rating, daily average financial asset amount, and the cost attributes include: personal loan balance at a certain point in time; The step of obtaining a financial attribute score of each customer in the customer group with purchase expectations according to the financial attribute of each customer includes: The financial attribute score of each customer is calculated according to the following financial attribute score calculation formula: ;in, Fa represents the financial attribute score; S i Indicates i A benefit attribute value; D i Indicates i Cost attribute value; n 1 Represents the total number of benefit-type attributes; n 2 Indicates the total number of cost-type attributes; The inherent attributes of the customer include at least one of the following: name, gender, age, education, height, weight, ID number, mobile phone number, occupation; The step of obtaining the inherent attribute score of each customer in the customer group with purchase expectations according to the inherent attribute of each customer includes: Calculate the score of each attribute value of each inherent attribute according to the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product; For each customer in the customer group with purchase expectations, the product of the scores of the attribute values ​​of each inherent attribute of the customer is taken as the inherent attribute score of the customer.

2. The method according to claim 1, characterized in that Before inputting the characteristic information of each customer in the original customer group into the pre-trained customer group pre-selection model, the method further includes: The characteristic information of each customer in the target customer group who has purchased the target financial product is used as a training label, and the characteristic information of each customer in the sample customer group is input into the original model to train the customer group pre-selection model, wherein the customers in the target customer group and the customers in the sample customer group are screened from the original customer group.

3. The method according to claim 1, characterized in that The customer's purchase behavior attributes include at least one of the following: the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product; The step of obtaining a purchase behavior attribute score of each customer in the customer group with purchase expectations according to the purchase behavior attribute of each customer includes: The purchase behavior attribute score of each customer is calculated according to the following purchase behavior attribute score formula: ;in, Be represents the purchase behavior attribute score; L i Indicates i Purchase behavior attribute value; n Indicates the total number of purchase behavior attributes.

4. The method according to claim 1, characterized in that: Determining the customer's expected purchase value of the target financial product according to the attribute score and the pre-selected score of each customer includes: The customer's expected purchase value for the target financial product is determined based on the pre-selected score, inherent attribute score and / or financial attribute score and / or purchase behavior attribute score of each customer.

5. The method according to claim 1, characterized in that After determining the customer's expected purchase value of the target financial product according to the pre-selected score and the attribute score of each customer, the method further includes: Determining a recommended customer group for the target financial product based on each customer's expected purchase value of the target financial product; The target financial product is marketed to the customers in the recommended customer group.

6. A financial product purchase prediction device, characterized in that: The device comprises: An acquisition module is used to input the characteristic information of each customer in the original customer group into a pre-trained customer group pre-selection model to obtain a customer group with purchase expectations for the target financial product and a pre-selection score for each customer in the customer group with purchase expectations; A first determination module, configured to determine an attribute score of each customer in the customer group with purchase expectations according to the attribute information of each customer; A second determination module, configured to determine the customer's expected purchase value of the target financial product based on the pre-selected score and the attribute score of each customer; The attributes of the customer include at least one of the following: inherent attributes, financial attributes, and purchase behavior attributes; the first determination module includes: a first acquisition unit, used to acquire the inherent attribute score of each customer in the customer group with purchase expectations according to the inherent attributes of each customer; a second acquisition unit, used to acquire the financial attribute score of each customer in the customer group with purchase expectations according to the financial attributes of each customer in the customer group with purchase expectations; and a third acquisition unit, used to acquire the purchase behavior attribute score of each customer in the customer group with purchase expectations according to the purchase behavior attributes of each customer; The inherent attributes of the customer include at least one of the following: name, gender, age, education, height, weight, ID number, mobile phone number, occupation; The first acquisition unit is specifically used for: Calculate the score of each attribute value of each inherent attribute according to the attribute value of each inherent attribute of each customer in the target customer group who has purchased the target financial product; For each customer in the customer group with purchase expectations, the product of the scores of the attribute values ​​of each inherent attribute of the customer is used as the inherent attribute score of the customer; The financial attributes of the customer include benefit attributes and cost attributes. The benefit attributes include at least one of the following: bank service star rating, daily average financial asset amount, and the cost attributes include: personal loan balance at a certain point in time; The second acquisition unit is specifically used for: The financial attribute score of each customer is calculated according to the following financial attribute score calculation formula: ;in, Fa represents the financial attribute score; S i Indicates i A benefit attribute value; D i Indicates i Cost attribute value; n 1 Represents the total number of benefit-type attributes; n 2 Indicates the total number of cost attributes.

7. The device according to claim 6, characterized in that The device also includes: The model training module is used to use the characteristic information of each customer in the target customer group who has purchased the target financial product as a training label, input the characteristic information of each customer in the sample customer group into the original model, and train to obtain the customer group pre-selection model, wherein the customers in the target customer group and the customers in the sample customer group are screened from the original customer group.

8. The device according to claim 6, characterized in that The customer's purchase behavior attributes include at least one of the following: the latest purchase time of the target financial product, the holding period of the target financial product, the total purchase amount of the target financial product, and the number of purchases of the target financial product; The third acquisition unit is specifically used for: The purchase behavior attribute score of each customer is calculated according to the following purchase behavior attribute score formula: ;in, Be represents the purchase behavior attribute score; L i Indicates i Purchase behavior attribute value; n Indicates the total number of purchase behavior attributes.

9. The device according to claim 6, characterized in that The second determining module is specifically used for: The customer's expected purchase value for the target financial product is determined based on the pre-selected score, inherent attribute score and / or financial attribute score and / or purchase behavior attribute score of each customer.

10. The device according to claim 6, characterized in that The device also includes: A third determination module is used to determine a recommended customer group for the target financial product according to the expected purchase value of each customer for the target financial product; A marketing module is used to market the target financial product to the customers in the recommended customer group.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

12. 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 method according to any one of claims 1 to 5 are implemented.

Images