Risk assessment method and device based on random forest, equipment and medium
Through the risk assessment method based on random forests, customer characteristics are deeply analyzed and risk indicators are screened out, which solves the deviation problems existing in financial institutions in customer risk assessment, and realizes accurate assessment of customer risks and the provision of personalized transactions.
Patent Information
- Application Number
- CN202411671796.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-20
- Publication Date
- 2025-03-14
AI Technical Summary
In the prior art, financial institutions lack accurate data analysis methods when evaluating customer risks, resulting in risk estimate deviations and inability to meet customer personalized needs.
A risk assessment method based on random forests is adopted to provide personalized trading methods by in-depth analysis of customer characteristics, customer risk indicators are screened out, and the pre-established risk level model is evaluated.
It realizes accurate assessment of customer risks, improves the reliability of transaction behavior, and enables financial institutions to more accurately locate potential customers and provide personalized services.
Smart Images
Figure CN119624645A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a risk assessment method, device, equipment and medium based on random forest. Background Art
[0002] In an economy, transactions occur between financial institutions and individuals. Before a financial institution conducts a transaction with an individual, it often needs to assess the risk of the customer to ensure the reliability of the transaction. However, in reality, financial institutions only rely on traditional general risk assessment models to assess the risk of their customers, and are unable to conduct targeted analysis based on the actual situation of each financial institution, which will lead to deviations in the estimation of customer risks, thereby causing risks in transactions with customers.
[0003] In addition, transactions between financial institutions and individuals can be diverse, but if financial institutions only rely on traditional general risk assessment models to assess customer risks, it will be difficult for them to determine their customers' personalized needs, resulting in them being unable to meet the diverse needs of different customer groups through customized transaction methods.
[0004] The main reason for the above problems is often due to insufficient data analysis of customers. Financial institutions fail to make full use of big data or advanced data analysis and fail to deeply explore customer characteristics. Therefore, the existing technology lacks a method that can accurately screen customer risk indicators and deeply assess customer risks, so as to provide personalized transaction methods for potential customers in a targeted manner and improve the reliability of transaction behaviors between financial institutions and individuals. Summary of the invention
[0005] To solve the above problems, the embodiments of the present application provide a risk assessment method, device, equipment and medium based on random forest, which can use random forest to perform in-depth analysis of customer characteristics and screen out customer risk indicators, and assess customer risks based on risk level models and using customer risk indicators, thereby guiding financial institutions to provide personalized transaction methods to potential customers and improve the reliability of transaction behaviors.
[0006] The present application embodiment adopts the following technical solutions:
[0007] In a first aspect, a risk assessment method based on random forest is provided, the method comprising:
[0008] Obtain the customer's customer characteristics and use random forest to determine the importance of each customer characteristic;
[0009] Filter out customer risk indicators from each customer feature based on importance;
[0010] Determine the risk value of each customer risk indicator based on the pre-established risk level model;
[0011] Determine the customer's risk level based on the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0012] Optionally, obtain the customer's customer features and use random forest to determine the importance of each customer feature, including:
[0013] Obtain customer characteristics of the customer;
[0014] Use random forest to determine the contribution of each customer feature to each node of each decision tree in the random forest;
[0015] According to the contribution, the importance of each customer feature in the random forest is determined.
[0016] Optionally, customer risk indicators are screened out from each customer characteristic based on importance, including:
[0017] Sort each customer feature by importance;
[0018] The customer characteristics that meet the preset number of items are screened out from high to low in order of importance as customer risk indicators.
[0019] Optionally, customer risk indicators are screened out from each customer characteristic based on importance, including:
[0020] Customer characteristics whose importance is higher than the preset importance threshold are screened out as customer risk indicators.
[0021] Optionally, a pre-established risk level model is determined by:
[0022] Get customer characteristics of all historical customers;
[0023] Determine the completion degree of all customer characteristics based on historical customer transaction results;
[0024] Assign risk values to all customer characteristics based on their degree of completion.
[0025] Optionally, based on a pre-established risk level model, determine the risk value of each customer risk indicator, including:
[0026] Filter out each customer’s risk indicators from all customer characteristics;
[0027] Determine the risk value corresponding to each customer risk indicator.
[0028] Optionally, the risk level of the customer is determined according to the risk value of each customer risk indicator and the importance of each customer risk indicator, including:
[0029] The risk value of each customer risk indicator is multiplied by the importance of the corresponding customer risk indicator to obtain a weighted value;
[0030] Add up the weighted values to get the customer's risk level.
[0031] In a second aspect, a risk assessment device based on random forest is provided, the device comprising:
[0032] An importance determination unit, used to obtain customer characteristics of customers and determine the importance of each customer characteristic using random forest;
[0033] A customer risk indicator determination unit is used to select customer risk indicators from various customer characteristics according to importance;
[0034] A risk value determination unit, used to determine the risk value of each customer risk indicator based on a pre-established risk level model;
[0035] The risk level determination unit is used to determine the risk level of the customer according to the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0036] In a third aspect, a computer device is provided, 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 above-mentioned random forest-based risk assessment method when executing the computer program.
[0037] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program, and when the computer program is instructed by a processor, the steps of the above-mentioned random forest-based risk assessment method are implemented.
[0038] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:
[0039] The random forest-based risk assessment method provided by the present application obtains the customer characteristics of the customer, uses random forest to determine the importance of each customer characteristic, selects the customer risk indicators from each customer characteristic according to the importance, determines the risk value of each customer risk indicator based on the pre-established risk level model, and determines the customer's risk level according to the risk value of each customer risk indicator and the importance of each customer risk indicator. The random forest-based risk assessment method provided by the present application uses random forest to conduct an in-depth analysis of customer characteristics, accurately selects the customer's personalized customer risk indicators through the importance of customer characteristics, and makes the subsequent assessment more targeted; determines the risk value of the customer risk indicator based on the risk level model, and obtains the overall rating of the customer risk level by combining the risk value and importance, which can make the customer's risk assessment results more accurate. The random forest-based risk assessment method provided by the present application, applied to financial institutions, can enable financial institutions to more accurately locate potential customers, provide personalized transaction methods, and improve the reliability of transaction behavior. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0041] Figure 1 A schematic diagram showing a process of a risk assessment method based on random forests according to an embodiment of the present application is shown;
[0042] Figure 2 A schematic diagram showing a process flow of a risk assessment method based on random forests according to another embodiment of the present application;
[0043] Figure 3 A schematic diagram showing the structure of a risk assessment device based on random forest according to an embodiment of the present application is shown;
[0044] Figure 4 A schematic diagram of the structure of a computer device according to an embodiment of the present application is shown. DETAILED DESCRIPTION
[0045] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0046] The idea of this application is to use random forest to analyze customer characteristics and screen out customer risk indicators that are personalized for customers; then use customer risk indicators to evaluate customer risk levels, thereby forming customer risk ratings. The risk levels evaluated by the above method are customer-specific and can be used to deeply explore potential customers, allowing financial institutions to provide individuals with personalized transaction methods and improve the reliability of transaction behavior.
[0047] Figure 1 The following is a schematic diagram showing a process flow of a risk assessment method based on random forests proposed in one embodiment of the present application. Figure 1 As shown, the method includes steps S110 to S140:
[0048] Step S110, obtaining customer characteristics of customers, and using random forest to determine the importance of each customer characteristic.
[0049] This application conducts in-depth analysis based on the unique customer characteristics of each customer, so it is necessary to first obtain the customer characteristics of the customer. Financial institutions can store a large amount of information data related to customers in their own customer databases and obtain the customer characteristics of customers through the customer databases. The customer databases of financial institutions themselves are often limited. When the customer databases of financial institutions themselves do not include the customer characteristics of the customers that need to be obtained or the customer characteristics are relatively few, financial institutions can also obtain the customer characteristics of customers by crawling public information on the public network.
[0050] The customer characteristics of the customer may include, but are not limited to, the customer's basic information, the customer's financial information, the customer's relationship information, etc. Among them, the customer's basic information may include, but are not limited to, the customer's name, age, education level, marital status, etc.; the customer's financial information may include, but are not limited to, the customer's deposits, loans, housing conditions, etc.; the customer's relationship information may include, but are not limited to, the customer's occupation, unit, social relations, etc.
[0051] In actual situations, the total number of customer characteristics (i.e., all customer characteristics) that financial institutions need to obtain is certain, but financial institutions may not be able to obtain all of them. Therefore, financial institutions can obtain as many customer characteristics as possible within the total number of items. Then, the number of customer characteristics of different customers may be different. For example, a financial institution obtains the name, age, deposit, and occupation characteristics of customer A through a customer database and / or a public network, and these characteristics are used as customer characteristics of customer A; a financial institution obtains the name, education level, deposit, loan, and unit characteristics of customer B through a customer database and / or a public network, and these characteristics are used as customer characteristics of customer B. In other words, a customer's customer characteristics are a subset of all customer characteristics, and a customer's customer characteristics are unique to that customer.
[0052] After obtaining the customer's customer features, random forest is used to determine the importance of each customer feature. Random forest is an ensemble learning method that can combine multiple decision trees and obtain the final prediction result by voting or averaging. Random forest can reduce the problem of overfitting of a single decision tree and improve the stability and accuracy of the prediction results. In addition, random forest does not require post selection, but automatically constructs decision trees, thus avoiding the subjective bias caused by manual feature selection.
[0053] Random forests can be used not only for prediction, but also for feature selection. Therefore, this application can use random forests to evaluate the importance of each customer feature in each node of each tree in the random forest, and then use the importance of each customer feature to determine the importance of the customer feature. Since different customers have different customer features, the importance of personalized customer features of different customers can be analyzed through random forests.
[0054] Step S120, screening out customer risk indicators from each customer feature according to the importance.
[0055] After using random forest to determine the importance of each customer feature, the customer features can be sorted using the importance index to determine which customer features are more important and which are less important. Customer features with higher importance have a greater impact on the customer risk assessment results, while customer features with lower importance have a smaller impact on the customer risk assessment results. Therefore, a portion of customer features with higher importance can be screened out as customer risk indicators for subsequent customer risk assessment.
[0056] Since the acquired customer features are different for each customer, the importance ranking of the customer features of each customer determined by the random forest is different, and the customer features with higher importance selected based on the importance ranking are different. In other words, the technical solution of the present application selects personalized customer features corresponding to each customer as the customer risk indicator of the customer.
[0057] In the screening process, some customer features can be selected as customer risk indicators by limiting the importance or limiting the number of customer features as customer risk indicators, but not limited to. For example, an importance threshold can be set. When the importance of a certain item in the customer features is higher than the importance threshold, the item is selected and retained as one of the customer risk indicators, and the other items in the customer features are similarly selected, thereby obtaining a subset of customer features with an importance higher than the importance threshold. The importance threshold needs to be flexibly set according to actual conditions. If the importance threshold is set too high, the importance of the customer features determined by the random forest may not be higher than the threshold, resulting in failure to select a suitable subset from the customer features for subsequent analysis; if the importance threshold is set too low, the number of customer features determined by the random forest with an importance higher than the threshold may be too large, resulting in excessive subsequent analysis and calculation. For another example, the number of customer features selected as customer risk indicators can be limited. After determining the importance of each customer feature by the random forest, the importance can be sorted from high to low or from low to high. A fixed number of subsets are intercepted from the customer feature with the highest importance as the customer risk indicator.
[0058] Step S130, determining the risk value of each customer risk indicator based on the pre-established risk level model.
[0059] This application analyzes the risk of customers based on the screened customer risk indicators. During the analysis, a pre-established risk level model is used. The risk level model can be established based on the customer risk indicators of historical customers and the transaction results of historical customers. The transaction results can include but are not limited to on-time repayment, overdue repayment, etc.
[0060] As an optional implementation method, the customer risk indicators corresponding to historical customers with different transaction results can be analyzed. The proportions of various customer risk indicators appearing under a certain transaction result are different, indicating that the influence of each customer risk indicator on this transaction result is different. A certain risk value is assigned to each customer risk indicator according to the different influences, so as to establish a risk level model based on each customer risk indicator and its corresponding risk value.
[0061] In the established risk level model, the number of items of all customer risk indicators is the number of items of all customer characteristics. In other words, each of all customer characteristics corresponds to a corresponding risk value. By bringing the customer risk indicators of the customer into the established risk level model, the risk values corresponding to each customer risk indicator of the customer can be determined as the risk values of each customer risk indicator.
[0062] As an optional implementation, the customer risk indicators of all historical customers and the transaction results of their corresponding historical customers can be used as samples, the customer risk indicators of a part of historical customers and the transaction results of their corresponding part of historical customers can be used as training sets, and the customer risk indicators of another part of historical customers and the transaction results of another part of historical customers can be used as test sets, and a risk level model can be trained through machine learning.
[0063] In the trained risk level model, the number of items of all customer risk indicators is the number of items of all customer characteristics. Since different transaction results can indicate different risks, such as low risk corresponding to on-time repayment and high risk corresponding to overdue repayment, corresponding risk values can be assigned to different transaction results. The trained risk level model can predict the transaction results corresponding to each of all customer characteristics, so the customer risk indicators of the customer are input into the trained risk level model to obtain the transaction results corresponding to the customer risk indicators of the customer, and then determine the risk values corresponding to the customer risk indicators of the customer as the risk values of each customer risk indicator.
[0064] Step S140, determining the risk level of the customer according to the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0065] Since the importance of each customer risk indicator of the customer has a certain impact on the final customer risk overall rating, this application, after determining the risk value of each customer risk indicator of the customer, comprehensively considers the importance of each customer risk indicator of the customer and determines the risk level of the customer.
[0066] The risk level of a customer may be determined by, but is not limited to, the following methods. The importance of each customer risk indicator may be multiplied by the corresponding risk value, and the result with the largest value may be selected as the customer's risk level. Alternatively, the importance of each customer risk indicator may be multiplied by the corresponding risk value, and the sum or average of the results may be used as the customer's risk level.
[0067] It should be noted that assessing a customer's risk level can not only determine the risk of the customer, but also compare it with the risk levels of other customers to assess the customer's risk level. Therefore, after determining the risk value and importance of each customer's risk indicator, no matter how the customer's risk level is determined, as long as the determination method is the same for all customers, the risk levels of different customers can be compared horizontally.
[0068] In addition, the customer characteristics of the same customer may change in different time periods. For example, the customer's deposits, loans, units and other information may change. Therefore, the customer characteristics of the customer can be obtained at intervals of specific time periods, and the risk level of the customer at different times can be determined by the above method of the present application. Based on the changes in the risk level of the same customer at different times, the risk level of the customer can be compared vertically.
[0069] A customer's risk level can be used directly to guide financial institutions in evaluating whether the customer is a potential transaction customer. The customer's risk level can also be used to guide financial institutions in evaluating whether the customer is a potential transaction customer by comparing it horizontally with the risk levels of other customers or vertically with the customer's previous risk levels.
[0070] Therefore, after the step of determining the risk level of the customer according to the risk value of each customer risk indicator and the importance of each customer risk indicator, the method may further include: comparing the risk level of the customer with a preset risk level threshold, if the risk level of the customer is higher than the preset risk level threshold, determining that the customer is high risk; if the risk level of the customer is not higher than the preset risk level threshold, determining that the customer is low risk. The method may further include: comparing the risk level of the customer with the average value of the risk levels of all customers, if the risk level of the customer is higher than the average value of the risk levels of all customers, determining that the customer is high risk; if the risk level of the customer is not higher than the average value of the risk levels of all customers, determining that the customer is low risk. The method may further include: comparing the risk level of the customer with the average value of the historical risk levels of the customer, if the risk level of the customer is higher than the average value of the historical risk levels of the customer, determining that the customer is high risk; if the risk level of the customer is not higher than the average value of the historical risk levels of the customer, determining that the customer is low risk.
[0071] from Figure 1 It can be seen from the method shown that the random forest-based risk assessment method provided by the present application obtains the customer characteristics of the customer, uses random forest to determine the importance of each customer characteristic, screens out customer risk indicators from each customer characteristic according to the importance, determines the risk value of each customer risk indicator based on a pre-established risk level model, and determines the customer's risk level according to the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0072] The random forest-based risk assessment method provided by this application uses random forests to conduct in-depth analysis of customer characteristics, accurately screens the customer's personalized customer risk indicators through the importance of customer characteristics, and makes subsequent assessments more targeted; the risk value of the customer risk indicator is determined based on the risk level model, and the overall rating of the customer risk level is obtained by combining the risk value and importance, which can make the customer's risk assessment results more accurate. The random forest-based risk assessment method provided by this application is applied to financial institutions, which can enable financial institutions to more accurately locate potential customers, provide personalized transaction methods, and improve the reliability of transaction behaviors.
[0073] In some optional embodiments, in the above method, step S110 obtains customer characteristics of the customer, and uses random forest to determine the importance of each customer characteristic, including: obtaining customer characteristics of the customer; using random forest to determine the contribution of each customer characteristic to each node of each decision tree of the random forest; and determining the importance of each customer characteristic in the random forest based on the contribution.
[0074] Customer characteristics are characteristic data that can reflect the customer's default risk. Financial institutions can obtain as many customer characteristics as possible through their own customer databases and public networks. The number of customer characteristics of a customer must not be more than the total number of all customer characteristics. Assuming that the total number of customer characteristics is N, a sample with a capacity of N is formed. This sample with a capacity of N is extracted N times with replacement, and 1 is extracted each time, eventually forming N samples. Use these N samples to train a decision tree as the sample at the root node of the decision tree.
[0075] Each customer feature has different attributes. For example, education level can have multiple attributes such as primary school, junior high school, high school, and university, and deposits can have multiple attributes such as 0-10,000 yuan, 10,000 yuan-100,000 yuan, 100,000 yuan-1 million yuan, and more than 1 million yuan. When one of the N samples has M attributes, each node of the decision tree needs to be split, so m attributes are randomly selected from the M attributes, and then one attribute is selected from the m attributes using information gain or other strategies as the split attribute of the node.
[0076] In the process of forming a decision tree, each node must be split according to the above steps until it cannot be split anymore. Therefore, a large number of decision trees are established based on the above steps to form a random forest.
[0077] After constructing the random forest, the importance of each feature in the random forest can be evaluated. For example, if a feature is used in a node in a decision tree, the Gini index can be used to represent the contribution of the feature to the node in the decision tree. The importance of each feature in the random forest can be the average value of the Gini index of all nodes in all decision trees.
[0078] In some optional embodiments, in the above method, step S120 selects customer risk indicators from each customer feature according to importance, including: sorting each customer feature according to importance; and selecting customer features that meet a preset number of items from high to low importance as customer risk indicators.
[0079] After obtaining the importance of each customer feature, the corresponding subset can be selected from the customer features according to the importance as the customer risk indicator. An optional method is to sort the customer features of the customer from high to low or from low to high in order of importance, and then select a preset number of items from high to low in order of importance as the customer risk indicator.
[0080] In some optional implementations, in the above steps, step S120 selects customer risk indicators from each customer feature according to importance, including: selecting customer features whose importance is higher than a preset importance threshold as customer risk indicators.
[0081] Another optional method is to directly use the importance of the customer's customer characteristics as a screening condition, and the customer characteristics whose importance is higher than a preset importance threshold constitute the customer risk index.
[0082] In some optional embodiments, in the above method, the pre-established risk level model in step S130 is determined by the following method: obtaining customer characteristics of all historical customers; determining the completion degree corresponding to all customer characteristics based on the transaction results of historical customers; and assigning risk values to all customer characteristics based on the completion degree.
[0083] In order to establish a risk level model based on the customer risk indicators and transaction results of historical customers, it is first necessary to obtain the customer characteristics of all historical customers. Historical customers are customers who have conducted transactions in financial institutions, and the customer characteristics of historical customers are already stored in the customer database of the financial institution itself. Financial institutions can fully obtain the customer characteristics of all historical customers through their own customer database.
[0084] It should be noted that since the establishment of the risk level model requires determining the risk values of all customer characteristics, in the process of establishing the risk level model, customer risk indicators are no longer screened, but a comprehensive analysis is conducted through the customer characteristics of all historical customers.
[0085] After a historical customer has made a transaction with a financial institution, the financial institution can determine the transaction results through his or her continuous or one-time repayment behavior. For example, the transaction results of timely one-time repayment and / or timely installment repayment can be determined as timely repayment, and the transaction results of non-repayment and / or non-time installment repayment and / or insufficient repayment amount can be determined as overdue repayment. For historical customers who repay in installments after taking out a loan, even if a historical customer has not made all repayments, as long as he or she can repay in installments on time, he or she can be determined as repaying on time. For a historical customer with multiple loans, as long as the transaction result of one of the loans can be determined as overdue repayment, the transaction result of the historical customer is determined as overdue repayment. In other words, all historical customers can determine a transaction result.
[0086] After obtaining the customer characteristics and transaction results of all historical customers, for any historical customer, a corresponding relationship between its customer characteristics and its transaction results can be generated. That is, for any historical customer, multiple customer characteristics correspond to one transaction result. Then, according to the transaction results, all historical customers can be classified, and the customer characteristics of historical customers in different categories can be obtained.
[0087] In one category, the customer characteristics of historical customers are repeated. Then, according to the degree of repetition of customer characteristics, the degree of influence of different customer characteristics on the transaction results can be determined. For example, in the transaction result of on-time repayment, the customer characteristics with the highest degree of repetition have the greatest positive impact on repayment, and the customer characteristics with gradually decreasing degree of repetition have a gradually decreasing positive impact on repayment. In the transaction result of overdue repayment, the customer characteristics with the highest degree of repetition have the greatest negative impact on repayment, and the customer characteristics with gradually decreasing degree of repetition have a gradually decreasing negative impact on repayment. The results of the impact of customer characteristics on completion under various transaction results are combined to determine the completion corresponding to all customer characteristics.
[0088] The higher the completion degree, the more likely the corresponding customer characteristics are to be able to repay on time, and the lower the completion degree, the more likely the corresponding customer characteristics are to be overdue. Therefore, the risk values from low to high can be assigned to customer characteristics according to the completion degree from high to low. The size of the assigned risk value can be determined according to the actual calculation requirements, as long as it meets the above allocation rules.
[0089] The following uses specific examples to explain in more detail how to determine the completion degree corresponding to all customer characteristics. For historical customers 1, 2, 3, 4, and 5, among them, 1, 2, and 3 repay on time, and 4 and 5 repay overdue. The customer characteristics of historical customer 1 are a, b, c, and d, the customer characteristics of historical customer 2 are b, d, and e, the customer characteristics of historical customer 3 are a, e, and f, the customer characteristics of historical customer 4 are c, d, and f, and the customer characteristics of historical customer 5 are b and f. Then, under the classification of transaction results of on-time repayment, a is repeated 2 times, b is repeated 2 times, c is repeated 1 time, d is repeated 2 times, e is repeated 2 times, and f is repeated 1 time. Under the classification of transaction results of overdue repayment, a is repeated 0 times, b is repeated 1 time, c is repeated 1 time, d is repeated 1 time, e is repeated 0 times, and f is repeated 2 times. The positive repayment degree can be obtained by taking the quotient of the number of repetitions of each customer feature under the on-time repayment category and the total number of customer features under the category, and the negative repayment degree can be obtained by taking the quotient of the number of repetitions of each customer feature under the overdue repayment category and the total number of customer features under the category. The positive repayment degree is subtracted from the negative repayment degree to determine the completion degree of the customer feature. That is, the completion degree of a is 0.2, the completion degree of b is 0, the completion degree of c is -0.1, the completion degree of d is 0, the completion degree of e is 0.2, and the completion degree of f is -0.1. Then, according to the above completion degree, a can be assigned a risk value of 10, b can be assigned a risk value of 50, c can be assigned a risk value of 75, d can be assigned a risk value of 50, e can be assigned a risk value of 10, and f can be assigned a risk value of 75. The above examples are only for illustrative purposes.
[0090] Since the transaction behavior of financial institutions is ongoing, the customer characteristics and transaction results of all historical customers tend to change dynamically. The risk level model established above is continuously optimized as the customer characteristics and transaction results of all historical customers change.
[0091] In some optional implementations, in the above method, step S130 determines the risk value of each customer risk indicator based on a pre-established risk level model, including: screening out each customer risk indicator from all customer characteristics; and determining the risk value corresponding to each customer risk indicator.
[0092] After the risk level model is established by the above method, each customer feature has its corresponding risk value. Then the customer risk index of the screened customers can always find the corresponding customer feature item from the above risk level model. After finding the customer feature item, the corresponding risk value corresponding to each customer risk index (and finding the corresponding customer feature item) is determined.
[0093] Continuing with the above specific example, the customer characteristics obtained are a, b, d, e, and f, and the customer risk indicators of the customer screened by random forest are b, e, and f. Then, after the customer risk indicators are brought into the above risk level model, it can be determined that the risk value corresponding to b is 50, the risk value corresponding to e is 10, and the risk value corresponding to f is 75.
[0094] In some optional embodiments, in the above method, step S140 determines the customer's risk level based on the risk value of each customer risk indicator and the importance of each customer risk indicator, including: multiplying the risk value of each customer risk indicator by the importance of the corresponding customer risk indicator to obtain a weighted value; adding each weighted value to obtain the customer's risk level.
[0095] After determining the risk value of each customer risk indicator, this application still comprehensively evaluates the customer's risk level by combining the importance of each customer risk indicator. That is, the risk value of a customer risk indicator of the customer is multiplied by the importance of the customer risk indicator in the random forest to obtain the weighted value of the customer risk indicator. Then, the weighted values of each customer risk indicator of the customer are added together to obtain the sum as the customer's risk level, that is, the customer's total risk rating.
[0096] Figure 2 A schematic diagram of a risk assessment method based on random forests proposed in another embodiment of the present application is shown. Figure 2 As shown, the method provided in this embodiment includes the following steps.
[0097] Step S201, obtaining customer characteristics of the customer.
[0098] Step S202, using random forest, determining the contribution of each customer feature in each node of each decision tree of the random forest; and determining the importance of each customer feature in the random forest according to the contribution.
[0099] Step S203, sorting each customer feature according to the importance; and selecting customer features that meet the preset number of items from high to low importance as customer risk indicators.
[0100] Alternatively, in step S204, customer characteristics whose importance is higher than a preset importance threshold are screened out as customer risk indicators.
[0101] Step S205, obtaining customer characteristics of all historical customers; determining the completion degree corresponding to all customer characteristics according to the transaction results of historical customers; and assigning risk values to all customer characteristics according to the completion degree.
[0102] Step S206, screening out each customer risk indicator from all customer characteristics; determining the risk value corresponding to each customer risk indicator.
[0103] Step S207, multiplying the risk value of each customer risk indicator by the importance of the corresponding customer risk indicator to obtain a weighted value; and adding the weighted values to obtain the customer's risk level.
[0104] Figure 3 A schematic diagram of the structure of a risk assessment device based on random forests according to an embodiment of the present application is shown. Figure 3 As shown, the device 300 includes:
[0105] An importance determination unit 310 is used to obtain customer features of customers and determine the importance of each customer feature using a random forest;
[0106] The customer risk indicator determination unit 320 is used to select the customer risk indicator from each customer feature according to the importance;
[0107] A risk value determination unit 330, for determining the risk value of each customer risk indicator based on a pre-established risk level model;
[0108] The risk level determination unit 340 is used to determine the risk level of the customer according to the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0109] In some optional embodiments, in the above-mentioned device 300, the importance determination unit 310 can be specifically used to: obtain customer characteristics of the customer; use random forest to determine the contribution of each customer characteristic to each node of each decision tree of the random forest; and determine the importance of each customer characteristic in the random forest based on the contribution.
[0110] In some optional implementations, in the above-mentioned device 300, the customer risk indicator determination unit 320 can be specifically used to: sort each customer feature according to importance; and filter out customer features that meet a preset number of items from high to low importance as customer risk indicators.
[0111] In some optional implementations, in the above-mentioned device 300, the customer risk indicator determination unit 320 can be specifically used to: screen out customer characteristics whose importance is higher than a preset importance threshold as customer risk indicators.
[0112] In some optional embodiments, in the above-mentioned device 300, the risk value determination unit 330 can be specifically used to: obtain customer characteristics of all historical customers; determine the completion degree corresponding to all customer characteristics based on the transaction results of historical customers; and assign risk values to all customer characteristics based on the completion degree.
[0113] In some optional implementations, in the above-mentioned device 300, the risk value determination unit 330 can also be specifically used to: screen out each customer risk indicator from all customer characteristics; and determine the risk value corresponding to each customer risk indicator.
[0114] In some optional implementations, in the above-mentioned device 300, the risk level determination unit 340 can be specifically used to: multiply the risk value of each customer risk indicator by the importance of the corresponding customer risk indicator to obtain a weighted value; and add each weighted value to obtain the customer's risk level.
[0115] It should be noted that the above-mentioned random forest-based risk assessment device 300 can implement the above-mentioned random forest-based risk assessment methods one by one, which will not be described in detail.
[0116] Figure 4 A schematic diagram showing the structure of a computer device according to an embodiment of the present application is shown. Figure 4 As shown, the internal structure of the computer device may include a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, the functions or steps of the risk assessment method based on random forests are implemented.
[0117] In one embodiment, the computer device provided by the present application includes a memory and a processor, the memory stores a database and a computer program that can be run on the processor, and the processor implements the following steps when executing the computer program:
[0118] Obtain the customer's customer characteristics and use random forest to determine the importance of each customer characteristic;
[0119] Filter out customer risk indicators from each customer feature based on importance;
[0120] Determine the risk value of each customer risk indicator based on the pre-established risk level model;
[0121] Determine the customer's risk level based on the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0122] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:
[0123] Obtain the customer's customer characteristics and use random forest to determine the importance of each customer characteristic;
[0124] Filter out customer risk indicators from each customer feature based on importance;
[0125] Determine the risk value of each customer risk indicator based on the pre-established risk level model;
[0126] Determine the customer's risk level based on the risk value of each customer risk indicator and the importance of each customer risk indicator.
[0127] It should be noted that the above functions or steps that can be implemented by the computer device or computer-readable storage medium can refer to the relevant description in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.
[0128] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0129] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0130] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A risk assessment method based on random forest, characterized in that: The method comprises: Obtain customer characteristics of the customer, and determine the importance of each of the customer characteristics using random forest; According to the importance, screening out customer risk indicators from each of the customer characteristics; Determine the risk value of each of the customer risk indicators based on a pre-established risk level model; The risk level of the customer is determined according to the risk value of each of the customer risk indicators and the importance of each of the customer risk indicators.
2. The random forest-based risk assessment method according to claim 1, characterized in that: The step of obtaining the customer characteristics of the customer and determining the importance of each of the customer characteristics using random forest includes: Obtain customer characteristics of the customer; Using random forest, determining the contribution of each of the customer features to each node of each decision tree of the random forest; According to the contribution, the importance of each of the customer features in the random forest is determined.
3. The random forest-based risk assessment method according to claim 1, characterized in that: The step of selecting customer risk indicators from each of the customer characteristics according to the importance includes: According to the importance, sorting each of the customer characteristics; According to the importance, the customer characteristics that meet the preset number of items are screened out from high to low as customer risk indicators.
4. The random forest-based risk assessment method according to claim 1, characterized in that: The step of selecting customer risk indicators from each of the customer characteristics according to the importance includes: The customer characteristics whose importance is higher than a preset importance threshold are screened out as customer risk indicators.
5. The random forest-based risk assessment method according to claim 1, characterized in that: The pre-established risk level model is determined by the following method: Get customer characteristics of all historical customers; Determine the completion degree corresponding to all the customer characteristics according to the transaction results of the historical customers; A risk value is assigned to all of the customer characteristics based on the degree of completion.
6. The random forest-based risk assessment method according to claim 5, characterized in that: Determining the risk value of each of the customer risk indicators based on the pre-established risk level model includes: Screening out each of the customer risk indicators from all of the customer characteristics; Determine the risk value corresponding to each of the customer risk indicators.
7. The random forest-based risk assessment method according to claim 1, characterized in that: Determining the risk level of the customer according to the risk value of each of the customer risk indicators and the importance of each of the customer risk indicators includes: Multiplying the risk value of each customer risk indicator by the importance of the corresponding customer risk indicator to obtain a weighted value; The weighted values are added together to obtain the risk level of the customer.
8. A risk assessment device based on random forest, characterized in that: The device comprises: An importance determination unit, used to obtain customer characteristics of the customer, and determine the importance of each of the customer characteristics using random forest; A customer risk indicator determination unit, configured to select a customer risk indicator from each of the customer characteristics according to the importance; A risk value determination unit, used to determine the risk value of each of the customer risk indicators based on a pre-established risk level model; The risk level determination unit is used to determine the risk level of the customer according to the risk value of each of the customer risk indicators and the importance of each of the customer risk indicators.
9. A computer 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 random forest-based risk assessment method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is instructed by a processor, it implements the steps of the random forest-based risk assessment method according to any one of claims 1 to 7.