Automatic customer screening method and system
Through automated customer screening methods, candidate customers are screened and influenced by multi-dimensional screening conditions, solving the problems of inefficient and low transparency of existing customer screening methods, and achieving more efficient and accurate customer screening and business response.
Patent Information
- Application Number
- CN202411717008.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-02-18
AI Technical Summary
Existing customer screening methods have problems such as inefficiency, low decision transparency and slow business response, resulting in potential customer churn and business opportunities.
The automated customer screening method is adopted to collect the candidate customer set and filter based on the initial multi-dimensional screening conditions to determine the screening pass rate. When the pass rate does not reach the preset range, the screening pass condition is graded and iteratively corrected based on the impact degree of each screening condition. Until the pass rate reaches the preset range.
It improves the efficiency and accuracy of customer screening, enhances the flexibility and adaptability of decision-making, reduces potential customer churn and business opportunities losses, and provides financial institutions with more accurate and reliable customer evaluation tools.
Smart Images
Figure CN119477521A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data screening, and in particular to an automated customer screening method and an automated customer screening system. Background Art
[0002] In the modern financial industry, customer loan approval and credit assessment are one of the core businesses. Financial institutions usually rely on multiple decision parameters (such as income level, credit score, borrowing history, etc.) to screen customers. However, existing screening methods often face many challenges, resulting in the loss of potential customers and missed business opportunities. First, despite the large number of decision parameters, the degree of influence of different parameters on customer approval rates varies significantly. For example, a high income level may have a greater impact on the loan application approval rate of some customers, while credit scores may be more important in other cases. This uneven impact makes the preset screening rules often unable to flexibly respond to actual conditions, resulting in failure to achieve the expected approval rate. Existing technologies usually rely on manual adjustment of screening rules, but this method has the problem of inefficiency. Manual adjustment is not only time-consuming, but also requires decision makers to have rich industry experience and keen insight, which is particularly difficult in a high-load business environment. In addition, the adjustment cycle is long, and decision makers need to review and revise the rules after each evaluation, which affects the response speed of the business.
[0003] The decision transparency of the manual adjustment process is low, and it is difficult to track and analyze the effect of each adjustment. This opacity may lead to biased or unfair treatment of customers, which will affect customer experience and institutional reputation. Customers may be confused because they cannot understand the reasons for being rejected, which will lead to a negative impression of financial institutions. Based on the above problems, the market urgently needs a new customer screening solution. Summary of the invention
[0004] The purpose of the embodiments of the present invention is to provide an automated customer screening method and system to at least solve the problem that existing customer screening solutions easily lead to loss of potential customers and missed business opportunities.
[0005] In order to achieve the above-mentioned purpose, the first aspect of the present invention provides an automated customer screening method, which includes: collecting a candidate customer set, and performing screening of the candidate customer set based on initial multidimensional screening conditions to obtain screened customers that meet all the initial multidimensional screening conditions; judging the screening pass rate of the screened customers that meet all the initial multidimensional screening conditions relative to the candidate customer set, and judging whether the screening pass rate is within a preset pass rate range; when the screening pass rate is not within the preset pass rate range, grading the degree of influence of each screening condition based on the customer pass situation of each screening condition; iteratively correcting the screening conditions based on the graded screening conditions, and re-executing the candidate customer set screening after each correction until the latest screening pass rate is within the preset pass rate range.
[0006] Optionally, the screening of the candidate customer set is performed based on the initialization multidimensional screening conditions to obtain screened customers that meet all the initialization multidimensional screening conditions, including: collecting screening target information of each customer in the candidate customer set, and classifying the screening target information of each customer based on each screening condition in the initialization multidimensional screening conditions to obtain a screening target information subset for each customer corresponding to each screening condition; screening is performed based on matching the screening target information subset of each screening condition with the corresponding screening condition in turn, and the next screening condition screening is performed for customers who pass the current screening condition until the last screening condition is completed; if any screening condition fails, the corresponding customer will be directly judged as failing the screening.
[0007] Optionally, when the screening pass rate is not within the preset pass rate range, the degree of influence of each screening condition is graded based on the number of customers passing each screening condition, including: when the screening pass rate is not within the preset pass rate range, each screening condition is separately performed based on the screening target information subset of each customer corresponding to each screening condition, and the number of customers screened out corresponding to each screening condition is counted; the screening rate of each screening condition is determined based on the proportion of the number of customers screened out corresponding to each screening condition in the candidate customer set; the screening rates of each screening condition are sorted from large to small, and the degree of influence of each screening condition is determined based on the sorting result; wherein, the larger the screening rate, the higher the degree of influence of the corresponding screening condition.
[0008] Optionally, the screening rate of each screening condition is sorted from large to small, and the influence level of each screening condition is determined based on the sorting result, including: based on a preset number of influence level levels, determining the proportion of the screening conditions under each influence level in the total screening conditions; based on the sorted screening conditions, determining the influence level of each screening condition in turn based on the proportion of the screening conditions under each influence level in the total screening conditions, and obtaining each screening condition under each influence level.
[0009] Optionally, the iterative filtering condition modification based on the graded filtering conditions includes: taking the filtering conditions with the lowest impact level as fixed filtering conditions that are not modified during the iteration process; determining the adjustment priority of each filtering condition based on the impact level of other filtering conditions; wherein, the higher the impact level, the higher the adjustment priority of the corresponding filtering condition; iteratively adjusting each filtering condition based on the adjustment priority of each filtering condition; wherein, at least one filtering condition is adjusted in each round of iteration.
[0010] Optionally, each screening condition is iteratively adjusted based on the adjustment priority of each screening condition, including: setting maximizing the screening pass rate as the goal, constructing an objective function, and randomly selecting a portion of customers for preliminary screening, and recording the preliminary screening pass rate; using Gaussian process regression to establish a proxy model of the objective function, and performing adjustments to each screening condition in turn based on the adjustment priority of each screening condition, and updating the proxy model after the adjustment; repeatedly performing screening and updating the model until the screening pass rate reaches the requirement within a preset range.
[0011] Optionally, each filtering condition is iteratively adjusted based on the adjustment priority of each filtering condition, including: setting the objective function as the screening pass rate, and calculating the gradient of each filtering condition parameter based on the differential objective function; wherein, the higher the adjustment priority, the earlier the corresponding filtering condition performs parameter adjustment; applying the gradient descent method to the filtering condition with the highest adjustment priority, updating its parameters, and re-executing the screening according to the updated parameters, calculating the new screening pass rate and recording the results; executing parameter updates for each filtering condition in turn based on the adjustment priority until the latest screening pass rate is within the preset pass rate range.
[0012] Optionally, the determination of the screening pass rate of the screened customers who meet all the initialization multi-dimensional screening conditions relative to the candidate customer set, and the determination of whether the screening pass rate is within a preset pass rate range, includes: taking the proportion of the screened customers who meet all the initialization multi-dimensional screening conditions in the candidate customer set as the screening pass rate; collecting the preset pass rate range set by the user, and determining whether the current screening pass rate is within the preset pass rate range.
[0013] The second aspect of the present invention provides an automated customer screening system, the system comprising: a collection unit, for collecting a candidate customer set, and performing screening of the candidate customer set based on initial multidimensional screening conditions to obtain screened customers that meet all the initial multidimensional screening conditions; a judgment unit, for judging the screening pass rate of the screened customers that meet all the initial multidimensional screening conditions relative to the candidate customer set, and judging whether the screening pass rate is within a preset pass rate range; a grading unit, for grading the degree of influence of each screening condition based on the customer pass situation of each screening condition when the screening pass rate is not within the preset pass rate range; a screening unit, for iteratively correcting the screening conditions based on the graded screening conditions, and re-executing the candidate customer set screening after each correction until the latest screening pass rate is within the preset pass rate range.
[0014] On the other hand, the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed on a computer, enables the computer to execute the above-mentioned automated customer screening method.
[0015] Through the above technical scheme, the scheme of the present invention realizes efficient screening and analysis of customers by collecting candidate customer sets and performing preliminary screening based on multi-dimensional screening conditions. First, the system can automatically determine the pass rate of customers who meet the screening conditions relative to the overall candidate set, thereby evaluating the effectiveness of the current screening rules. If the pass rate does not reach the preset range, the system will further grade the degree of influence of each screening condition to ensure that the conditions with a greater impact on the pass rate are given priority. Then, by iteratively correcting these screening conditions, the system can dynamically optimize the screening rules and gradually improve the customer pass rate. Ultimately, this method not only improves the screening efficiency, but also enhances the flexibility and adaptability of decision-making, thereby effectively reducing the loss of potential customers and business opportunities, and providing financial institutions with a more accurate and reliable customer evaluation tool.
[0016] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:
[0018] Figure 1 is a flowchart of the steps of an automated customer screening method provided by one embodiment of the present invention;
[0019] Figure 2 It is a system structure diagram of an automated customer screening system provided by one embodiment of the present invention. DETAILED DESCRIPTION
[0020] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.
[0021] Figure 1 FIG. 1 is a flow chart of an automated customer screening method provided by an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides an automated customer screening method, the method comprising:
[0022] Step S10: Collecting a candidate customer set, and performing screening of the candidate customer set based on the initial multi-dimensional screening conditions, to obtain screened customers that meet all the initial multi-dimensional screening conditions.
[0023] Specifically, the screening target information of each customer in the candidate customer set is collected, and the screening target information of each customer is classified based on each screening condition in the initialized multi-dimensional screening conditions to obtain a screening target information subset corresponding to each screening condition for each customer; screening is performed in turn based on the matching of the screening target information subset of each screening condition with the corresponding screening condition, and the next screening condition is performed for customers who pass the current screening condition until the last screening condition is completed; if any screening condition fails, the corresponding customer will be directly judged as failing the screening.
[0024] In an embodiment of the present invention, the screening target information of each customer in the candidate customer set is collected in detail, including but not limited to the customer's income level, credit score, borrowing history and other relevant financial data. Through this information, the system can construct a subset of the screening target information of each customer under each screening condition. The key to this step is the comprehensiveness and accuracy of the data. Only by fully collecting the key information of the customer can a reliable basis be provided for subsequent screening. Each screening condition will be applied in turn and matched with the screening target information subset of the customer to determine whether each customer meets the specific screening condition. In this process, the screening is carried out layer by layer, that is, only when the customer passes the current screening condition, the system will continue to perform the screening of the next condition on it. This step-by-step screening method helps to exclude unqualified customers layer by layer, ensuring that the final selected customers are more potential and reliable.
[0025] Furthermore, the system sets a rule that if any of the screening conditions are not passed, the corresponding customer will be directly judged as not passing the screening. This strategy effectively avoids unnecessary subsequent evaluation work and improves the overall screening efficiency. Through this efficient screening method, financial institutions can identify qualified customers in a shorter period of time, thereby speeding up the loan approval and credit assessment process.
[0026] Based on the solution of the present invention, through this technical solution, financial institutions can significantly improve the efficiency and accuracy of customer screening. First, systematic data collection and classification make the screening process more transparent and controllable, and the decision-making basis is clear and definite. Secondly, the layer-by-layer screening method reduces the complexity of customer evaluation and avoids screening errors caused by complex conditions. At the same time, unqualified customers are quickly identified and excluded, reducing potential credit risks. In addition, the automated process based on multi-dimensional screening conditions allows financial institutions to adjust screening criteria more flexibly to adapt to market changes and fluctuations in customer demand, thereby improving the response speed and competitiveness of the business. This innovative method not only improves customer satisfaction, but also effectively reduces business losses caused by customer churn, bringing greater market opportunities to financial institutions. In general, while improving the efficiency of customer screening, this technical solution also provides important support for the sustainable development of financial services.
[0027] In a possible implementation, the multidimensional screening conditions include:
[0028] 1) Customer income (e.g. monthly income ≥ RMB 6,000);
[0029] 2) Repayment ability (e.g. repayment amount to income ratio ≤ 35%);
[0030] 3) Credit score (e.g. credit score ≥ 700 points);
[0031] 4) Liabilities ratio (e.g. liabilities to total assets ratio ≤ 50%);
[0032] 5) Historical transaction frequency (e.g., number of transactions per year ≥ 10).
[0033] Step S20: Determine the screening pass rate of customers who meet all the initialization multi-dimensional screening conditions relative to the candidate customer set, and determine whether the screening pass rate is within a preset pass rate range.
[0034] Specifically, the proportion of customers who meet all the initial multi-dimensional screening conditions in the candidate customer set is taken as the screening pass rate; the preset pass rate range set by the user is collected, and it is determined whether the current screening pass rate is within the preset pass rate range.
[0035] In an embodiment of the present invention, the system first performs multi-dimensional screening on all candidate customers to identify those customers that meet the set conditions. For example, assuming that there are 1,000 customers in the candidate customer set, of which 600 customers meet all the initial screening conditions, the screening pass rate is 60%. Then, the system compares this pass rate with the pass rate range preset by the user. The user can set a suitable pass rate range according to market conditions and business needs, for example, between 70% and 80%. When the system detects that the current screening pass rate is lower than the preset range, it will issue an alarm to prompt the decision maker to make adjustments. At this time, the decision maker can deeply analyze the reasons for the insufficient pass rate and optimize the screening rules. This data-driven analysis method can help organizations identify overly strict or unreasonable screening conditions, so as to make timely adjustments.
[0036] In one possible implementation, suppose that the customer screening system of a financial institution initially sets the screening rules as follows: the customer's monthly income must be no less than 6,000 yuan, the credit score must be no less than 700 points, and the debt ratio must not exceed 50%. After screening 1,000 customers, the system found that only 500 customers passed the screening, with an actual pass rate of 50%, far lower than the expected 70%. This result shows that the existing screening rules are too strict and fail to effectively cover more potential customers.
[0037] Step S30: When the screening pass rate is not within the preset pass rate range, the impact degree of each screening condition is graded based on the customer pass status of each screening condition.
[0038] Specifically, when the screening pass rate is not within the preset pass rate range, each screening condition is screened separately based on the screening target information subset corresponding to each screening condition of each customer, and the number of customers screened out corresponding to each screening condition is counted; the screening rate of each screening condition is determined based on the proportion of the number of customers screened out corresponding to each screening condition in the candidate customer set; the screening rates of each screening condition are sorted from large to small, and the influence level of each screening condition is determined based on the sorting results; wherein, the larger the screening rate, the higher the influence level of the corresponding screening condition.
[0039] Furthermore, the screening rate of each screening condition is sorted from large to small, and the influence level of each screening condition is determined based on the sorting result, including: based on a preset number of influence level levels, determining the proportion of the screening conditions under each influence level in the total screening conditions; based on the sorted screening conditions, the influence level of each screening condition is determined in turn based on the proportion of the screening conditions under each influence level in the total screening conditions, to obtain each screening condition under each influence level.
[0040] In the embodiment of the present invention, the screening rate reflects the actual impact of a certain screening condition on the customer pass rate. The higher the screening rate, the greater the impact of the condition on customer rejection. In order to more effectively optimize the screening rules, the system sorts the screening rates of all conditions from large to small, thereby determining the level of influence of each condition. This grading method allows decision makers to quickly identify the key factors that affect the pass rate and adjust the screening rules accordingly.
[0041] Furthermore, the system can define the proportion of each condition in the overall screening conditions based on the preset number of impact levels. Based on the sorting results, the system can classify the screening conditions into high, medium, and low impact levels, thereby providing guidance for subsequent rule adjustments. For example, if the screening rate of a certain screening condition is significantly higher than that of other conditions, decision makers should give priority to optimizing this condition.
[0042] Based on the solution of the present invention, firstly, the scientificity and transparency of the screening rules are improved, so that the optimization decision can be based on evidence; secondly, through real-time feedback, the system can quickly adapt to market changes and improve the customer approval rate; finally, customer satisfaction is enhanced and the loss of potential customers caused by unreasonable screening conditions is reduced.
[0043] Step S40: iteratively modify the screening conditions based on the classified screening conditions, and re-execute the screening of the candidate customer set after each modification until the latest screening pass rate is within the preset pass rate range.
[0044] Specifically, each filtering condition with the lowest impact level is used as a fixed filtering condition that is not modified during the iteration process; based on the impact level of other filtering conditions, the adjustment priority of each filtering condition is determined; wherein, the higher the impact level, the higher the adjustment priority of the corresponding filtering condition; each filtering condition is iteratively adjusted based on the adjustment priority of each filtering condition; wherein, at least one filtering condition is adjusted in each round of iteration.
[0045] Embodiment 1:
[0046] Execute adjustments to each screening condition based on Bayesian optimization, including: set maximizing the screening pass rate as the goal, construct an objective function, randomly select a portion of customers for preliminary screening, and record the preliminary screening pass rate; use Gaussian process regression to establish a proxy model of the objective function, and execute adjustments to each screening condition in turn based on the adjustment priority of each screening condition, and update the proxy model after the adjustment; repeat the screening and updating the model until the screening pass rate reaches the requirement within a preset range.
[0047] In this embodiment, the system sets a goal: to maximize the screening pass rate of customers. In order to achieve this goal, the system constructs an objective function, which is specifically a mathematical expression of the customer pass rate. Then, the system randomly selects a part of customers for preliminary screening and records the preliminary screening pass rate. Through preliminary data collection, the system can establish a proxy model based on Gaussian process regression, which can describe the behavior of the objective function in the parameter space. After the model is established, the system begins to adjust the conditions in turn according to the adjustment priority of each screening condition. Each time the adjustment is made, the system will screen based on the results predicted by the proxy model and update the model parameters. This process will be repeated until the screening pass rate reaches the requirements within the preset range. In this process, Bayesian optimization allows the system to continuously learn and update the model when adjusting the screening conditions, so that each adjustment is based on the latest data analysis and feedback. The advantage of this method is that it can efficiently explore the parameter space, so as to quickly find the best parameter combination that can improve the customer pass rate. Finally, through this method, the screening pass rate of financial institutions has increased to 80%, effectively expanding the customer base and enhancing market competitiveness.
[0048] Embodiment 2:
[0049] Execute adjustment of each screening condition based on the gradient descent method, including: setting the objective function as the screening pass rate, and calculating the gradient of each screening condition parameter based on the differential objective function; wherein, the higher the adjustment priority, the earlier the corresponding screening condition will be adjusted; apply the gradient descent method to the screening condition with the highest adjustment priority, update its parameters, and re-execute the screening according to the updated parameters, calculate the new screening pass rate and record the results; execute parameter updates for each screening condition in turn based on the adjustment priority until the latest screening pass rate is within the preset pass rate range.
[0050] In this embodiment, the system sets the screening pass rate as the objective function, differentiates the objective function, and calculates the gradient of each screening condition parameter. At this time, the system will adjust the parameters of the conditions with the highest degree of influence according to the preset adjustment priority. For example, suppose that the current impact level classification shows that "customer income" and "credit score" have the greatest impact on the pass rate. The system first applies the gradient descent method to adjust the "customer income" condition, and determines how to update the parameter to maximize the customer's pass rate by calculating the corresponding gradient. Subsequently, the system re-executes the screening of the candidate customer set based on the updated "customer income" threshold and calculates the new screening pass rate. After recording the new pass rate, the system will calculate the parameter updates of other screening conditions again and adjust them in turn according to the priority. This process will continue until the latest screening pass rate reaches the preset range.
[0051] In one possible implementation, in a customer screening system of a financial institution, the results of the preliminary screening show that the pass rate is only 50%, far below the preset 70%. In order to improve this pass rate, the system decides to conduct a hierarchical analysis of the degree of influence of the screening conditions. The system organizes and analyzes the candidate customer data. Assume that there are 1,000 customers in the candidate customer set, of which 600 customers fail to pass the screening. The system analyzes the impact of each screening condition on the customer one by one. Taking "customer income" as an example, it is found that among the 500 customers who failed, 450 have a monthly income between 5,000 yuan and 6,000 yuan. Based on this finding, the system sets the impact of the income screening condition to the highest.
[0052] Next, the system conducted similar analyses on other screening conditions. For "credit score," the system found that 330 rejected customers had credit scores below 700, indicating that this threshold also needed to be adjusted. The screening condition for debt ratio showed that although 200 of the rejected customers had a debt ratio of more than 50%, the impact on the pass rate was relatively small, so the adjustment priority of this condition was lower.
[0053] After detailed statistics and analysis, the system generated a customer distribution analysis report, which included the percentage of customers who passed and rejected each screening parameter, and the impact analysis on the pass rate. Finally, the system proposed the following adjustment suggestions:
[0054] 1) Adjustment of customer income threshold: The “monthly income” requirement will be lowered from RMB 6,000 to RMB 5,500 to cover more potential customers.
[0055] 2) Adjustment of credit score threshold: Lower the credit score requirement from 700 to 650 to attract more customers to apply.
[0056] 3) Repayment ability and debt ratio: It is recommended to make slight adjustments to these two conditions to optimize the screening effect.
[0057] 4) Transaction frequency remains unchanged: Since it has the least impact on the approval rate, the current standard is maintained.
[0058] By implementing these adjustments, the financial institution was able to reapply the screening rules and rescreen 1,000 customers. The results showed that the pass rate increased to 72%, successfully achieving the preset target, and a wider range of potential customers were absorbed.
[0059] In another possible implementation:
[0060] 1) Adjustment of customer income threshold: The initial income requirement was no less than 6,000 yuan per month, but the system found that many customers' income was between 5,000 and 6,000 yuan. Therefore, the system lowered the income threshold from 6,000 yuan to 5,500 yuan and re-screened. Pass rate before adjustment: 50%. Pass rate after adjustment: 60% (the number of customers who passed increased from 500 to 600).
[0061] 2) Credit score threshold adjustment: The initial credit score requirement was no less than 700 points. The system found that many customers had credit scores between 680 and 700 points. Therefore, the system lowered the credit score threshold from 700 to 680 points. Pass rate before adjustment: 60%. Pass rate after adjustment: 67% (the number of customers passing increased from 600 to 670). After this adjustment, the system re-screened and the pass rate increased from 50% to 67%.
[0062] After each parameter adjustment, the system automatically re-screens customers and calculates a new pass rate. Through multiple rounds of iterative adjustments, the system continuously optimizes parameters and gradually approaches the expected customer pass rate target (such as 70%). During each iteration, the system will provide real-time feedback based on the screening results and make further adjustments based on the parameter optimization algorithm.
[0063] Example optimization: The system further found that a moderate adjustment of repayment capacity (from 35% to 40%) would have a positive impact on the pass rate. After the adjustment, the pass rate increased from 67% to 69%. Then, the system introduced alternative conditions, such as if the customer's credit score was low but the income was higher than 7,000 yuan, the credit score requirements could be relaxed, which further increased the pass rate to 71%.
[0064] Preferably, when the screening results reach the expected target, as shown in Table 1, the system generates an optimization report, including the changes before and after the adjustment of each parameter, the improvement of the customer pass rate, and the visualization of the final screening results. In addition, the system will automatically record the historical version of each adjustment to ensure that the optimization process of the screening rules is transparent and traceable. The reports output by the system include but are not limited to:
[0065] 1) Comparison table of pass rates before and after parameter adjustment;
[0066] 2) The adjustment range of each parameter threshold and its impact on the pass rate;
[0067] 3) Customer distribution changes during the screening process;
[0068] 4) Parameter optimization log automatically generated by the system.
[0069] Table 1 Example report
[0070] Parameter name Initial Threshold Adjusted Threshold Pass rate before adjustment Adjusted pass rate Customer income level ≥6000 yuan ≥5500 yuan 50％ 60％ Credit score ≥700 points ≥680 points 60％ 67％ Repayment ability ≤35％ ≤40％ 67％ 69％ Debt ratio ≤50％ ≤50％ 69％ 69％ Alternative conditions none Introducing high income compensation 69％ 71％
[0071] Based on the solution of the present invention, the system can automatically analyze the impact of each parameter on the pass rate without manual intervention, and dynamically adjust the screening threshold based on customer distribution to optimize the screening rules. Through the streaming data processing architecture, the system can achieve real-time optimization feedback of screening rules under large-scale customer data to ensure rapid response. The system uses advanced parameter optimization algorithms (such as Bayesian optimization) for iterative adjustments, which can accurately find the key factors affecting the pass rate and quickly make corresponding adjustments to achieve the expected goals. Through automated screening rule optimization, the system maintains efficient screening capabilities in large-scale data scenarios and improves overall screening efficiency. The system automatically generates optimization logs and screening reports to ensure that screening decisions are transparent and reviewable, which is convenient for the business team to supervise and trace back.
[0072] Figure 2 FIG. 1 is a system structure diagram of an automated customer screening system provided by an embodiment of the present invention. Figure 2 As shown, an embodiment of the present invention provides an automated customer screening system, the system comprising: a collection unit, used to collect a candidate customer set, and perform screening of the candidate customer set based on an initialization multi-dimensional screening condition to obtain screened customers that meet all the initialization multi-dimensional screening conditions; a judgment unit, used to judge the screening pass rate of the screened customers that meet all the initialization multi-dimensional screening conditions relative to the candidate customer set, and judge whether the screening pass rate is within a preset pass rate range; a grading unit, used to grade the degree of influence of each screening condition based on the customer pass situation of each screening condition when the screening pass rate is not within the preset pass rate range; a screening unit, used to iteratively modify the screening conditions based on the graded screening conditions, and re-execute the candidate customer set screening after each modification until the latest screening pass rate is within the preset pass rate range.
[0073] The embodiment of the present invention further provides a computer-readable storage medium, on which instructions are stored, which, when executed on a computer, enable the computer to execute the above-mentioned automatic customer screening method.
[0074] Those skilled in the art will understand that all or part of the steps in the method for implementing the above-mentioned embodiments can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including several instructions for making a single-chip microcomputer, a chip or a processor (processor) 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.
[0075] The optional embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the technical concept of the embodiments of the present invention, the technical scheme of the embodiments of the present invention can be subjected to a variety of simple modifications, and these simple modifications all belong to the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the embodiments of the present invention will not further describe various possible combinations.
[0076] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the embodiments of the present invention, they should also be regarded as the contents disclosed in the embodiments of the present invention.
Claims
1. An automated customer screening method, characterized in that: The method comprises: Collecting a candidate customer set, and performing screening of the candidate customer set based on the initial multi-dimensional screening conditions, to obtain screening customers that meet all the initial multi-dimensional screening conditions; Determine the screening pass rate of customers who meet all the initial multi-dimensional screening conditions relative to the candidate customer set, and determine whether the screening pass rate is within a preset pass rate range; When the screening pass rate is not within the preset pass rate range, the impact of each screening condition is graded based on the number of customers passing each screening condition; The screening conditions are iteratively modified based on the graded screening conditions, and the candidate customer set screening is re-executed after each modification until the latest screening pass rate is within the preset pass rate range.
2. The method according to claim 1, characterized in that The step of performing the screening of the candidate customer set based on the initialization multi-dimensional screening conditions to obtain the screened customers that meet all the initialization multi-dimensional screening conditions includes: Collecting screening target information of each customer in the candidate customer set, and classifying the screening target information of each customer based on each screening condition in the initial multi-dimensional screening condition, to obtain a screening target information subset of each customer corresponding to each screening condition; Filtering is performed based on the matching of the filter target information subset of each filter condition with the corresponding filter condition in turn, and the next filter condition is performed for customers who pass the current filter condition until the last filter condition is completed; If any of the screening conditions is not met, the corresponding customer will be directly judged as having failed the screening.
3. The method according to claim 2, characterized in that When the screening pass rate is not within the preset pass rate range, the impact of each screening condition is graded based on the number of customers passing each screening condition, including: When the screening pass rate is not within the preset pass rate range, each screening condition is screened separately based on the screening target information subset corresponding to each screening condition of each customer, and the number of customers screened out corresponding to each screening condition is counted; Determine the screening rate of each screening condition based on the proportion of the number of customers screened out corresponding to each screening condition in the candidate customer set; The screening rate of each screening condition is sorted from large to small, and the influence level of each screening condition is determined based on the sorting result; wherein, The greater the screening rate, the higher the level of influence of the corresponding screening condition.
4. The method according to claim 3, characterized in that The screening rate of each screening condition is sorted from large to small, and the influence level of each screening condition is determined based on the sorting result, including: Based on the number of preset impact levels, determine the proportion of screening conditions at each impact level in the total screening conditions; Based on the sorted screening conditions, the impact level of each screening condition is determined in turn based on the proportion of the screening conditions at each impact level in the total screening conditions, and the screening conditions at each impact level are obtained.
5. The method according to claim 4, characterized in that The iteratively modifying the screening conditions based on the classified screening conditions includes: The screening conditions with the lowest impact level are used as fixed screening conditions that are not modified during the iteration process; Based on the impact level of other screening conditions, determine the adjustment priority of each screening condition; among them, The higher the impact level, the higher the adjustment priority of the corresponding filtering conditions; Iteratively adjust each filtering condition based on the adjustment priority of each filtering condition; wherein, At least one filter condition is adjusted in each iteration.
6. The method according to claim 5, characterized in that Iteratively adjust each filter condition based on the adjustment priority of each filter condition, including: Set the goal of maximizing the screening pass rate, construct the objective function, randomly select a portion of customers for preliminary screening, and record the preliminary screening pass rate; Using Gaussian process regression to establish a proxy model of the objective function, and adjusting each screening condition in turn based on the adjustment priority of each screening condition, and updating the proxy model after the adjustment; The screening and model updating are repeated until the screening pass rate reaches the requirement within the preset range.
7. The method according to claim 5, characterized in that Iteratively adjust each filter condition based on the adjustment priority of each filter condition, including: The objective function is set as the screening pass rate, and the gradient of each screening condition parameter is calculated based on the differential objective function; where: The higher the adjustment priority, the earlier the corresponding filter conditions will be adjusted; Apply the gradient descent method to the screening condition with the highest adjustment priority, update its parameters, re-perform the screening based on the updated parameters, calculate the new screening pass rate and record the results; Based on the adjustment priority, the parameters of each screening condition are updated in turn until the latest screening pass rate is within the preset pass rate range.
8. The method according to claim 1, characterized in that The determining of the screening pass rate of the screening customers that meet all the initialization multi-dimensional screening conditions relative to the candidate customer set, and determining whether the screening pass rate is within a preset pass rate range, includes: The proportion of customers who meet all the initial multi-dimensional screening conditions in the candidate customer set is taken as the screening pass rate; Collect the preset pass rate range set by the user and determine whether the current screening pass rate is within the preset pass rate range.
9. An automated customer screening system, characterized in that: The system comprises: A collection unit, used for collecting a candidate customer set, and performing screening of the candidate customer set based on the initialization multi-dimensional screening conditions, to obtain screening customers that meet all the initialization multi-dimensional screening conditions; A judging unit, used to judge the screening pass rate of the customers who meet all the initialization multi-dimensional screening conditions relative to the candidate customer set, and to judge whether the screening pass rate is within a preset pass rate range; A grading unit, for grading the degree of influence of each screening condition based on the customer passing status of each screening condition when the screening pass rate is not within a preset pass rate range; The screening unit is used to iteratively modify the screening conditions based on the graded screening conditions, and re-execute the screening of the candidate customer set after each modification until the latest screening pass rate is within a preset pass rate range.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the automated customer screening method described in any one of claims 1 to 8.