Method and device for evaluating difficulty level of payment reminding, and readable storage medium

Abstract

The invention provides a payment urging difficulty level evaluation method and device and a readable storage medium. The method comprises the following steps: acquiring payment urging evaluation data of an overdue customer; constructing overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers according to the payment reminding evaluation data of the overdue customers; the overdue behavior features, the income and liability features, the business cooperation features and the industry and commerce features of the overdue customers are input into a trained payment urging difficulty model, the payment urging difficulty degree score of the overdue customers is obtained, and the payment urging difficulty model is a model constructed based on a logistic regression algorithm. According to the method, the device and the readable storage medium, the problem that the evaluation result is one-sided and the real payment urging condition cannot be comprehensively reflected due to the single evaluation dimension of the existing payment urging difficulty level evaluation method can be solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device and readable storage medium for assessing the difficulty of payment collection. Background Art

[0002] Currently, government and enterprise clients have a large number of accounts receivable generated by overdue users, and they urgently need a scientific and quantified collection strategy to recover the accounts receivable. To this end, we can evaluate the difficulty of collecting payments from government and enterprise clients to predict the possibility of collecting accounts receivable from government and enterprise clients in the future, thereby optimizing the collection process, improving collection efficiency, and improving customer relationships.

[0003] However, the existing methods for assessing the difficulty of payment collection have a single assessment dimension, which leads to one-sided assessment results and cannot fully reflect the actual payment collection situation. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method, device and readable storage medium for assessing the difficulty of payment collection in view of the above-mentioned shortcomings of the prior art, so as to solve the problem that the existing method for assessing the difficulty of payment collection has a single assessment dimension, resulting in one-sided assessment results and unable to fully reflect the actual payment collection situation.

[0005] In a first aspect, the present invention provides a method for assessing the difficulty of payment collection, comprising:

[0006] Obtain collection assessment data for overdue customers;

[0007] Based on the collection assessment data of the overdue customers, construct the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers;

[0008] The overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, wherein the collection difficulty model is a model built based on a logistic regression algorithm.

[0009] Furthermore, before inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes:

[0010] According to the actual business scenario, select the historical collection assessment data of multiple collection customers from the historical data as the modeling sample data;

[0011] According to the modeling sample data, the target variable of the collection difficulty model is determined, and the prediction variables of the collection difficulty model are constructed, wherein the prediction variables include overdue behavior characteristics, income and liability characteristics, business cooperation characteristics, and industrial and commercial characteristics;

[0012] The collection difficulty model is constructed by using a logistic regression algorithm, and the collection difficulty model is trained based on the modeling sample data and the target variable and the predictor variable of the collection difficulty model to obtain the trained collection difficulty model.

[0013] Furthermore, determining the target variable of the collection difficulty model based on the modeling sample data specifically includes:

[0014] For each collection demand customer in the modeling sample data, the cumulative monthly repayment rate of the collection demand customer is calculated based on the monthly repayment amount and receivable amount in the historical collection demand evaluation data of the collection demand customer.

[0015] Calculate the index average repayment rate of the demanded customers based on the cumulative monthly repayment rate of the demanded customers each month;

[0016] Setting a target threshold based on the distribution of indexed average repayment rates of all the customers called for repayment;

[0017] For each of the collection demanding customers, whether the collection demanding customer is an easy-to-collect customer is determined based on the index average repayment rate of the collection demanding customer and the target threshold, and whether the collection demanding customer is an easy-to-collect customer is used as the target variable.

[0018] Further, judging whether the demanding customer is an easy-to-demand customer based on the index average repayment rate of the demanding customer and the target threshold value specifically includes:

[0019] If the index average repayment rate of the collection customer is greater than or equal to the target threshold, the collection customer is judged to be an easy-to-collect customer;

[0020] If the index average repayment rate of the collection customer is less than the target threshold, the collection customer is determined to be a difficult-to-collect customer.

[0021] Furthermore, before inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes:

[0022] A scorecard is constructed according to the training results of the trained collection difficulty model, wherein the scoring dimensions of the scorecard include income and liability dimension, business cooperation dimension, overdue behavior dimension and industrial and commercial characteristics dimension.

[0023] Furthermore, the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, which specifically includes:

[0024] Inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model, and scoring the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer based on the scoring dimensions of the scoring card by the trained collection difficulty model to obtain the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristic score of the overdue customer; and

[0025] The difficulty score of collecting payment from the overdue customer is calculated based on the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristics score of the overdue customer.

[0026] Furthermore, after inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes:

[0027] Based on the difficulty level of collection for overdue customers, corresponding measures will be taken to collect the payment.

[0028] In a second aspect, the present invention provides a device for assessing the degree of difficulty of payment collection, comprising:

[0029] The acquisition module is used to obtain the collection assessment data of overdue customers;

[0030] A construction module, connected to the acquisition module, for constructing overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers based on the collection assessment data of the overdue customers;

[0031] The evaluation module is connected to the construction module and is used to input the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers into the trained collection difficulty model to obtain the collection difficulty score of the overdue customers, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm.

[0032] In a third aspect, the present invention provides a device for assessing the difficulty of payment collection, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the method for assessing the difficulty of payment collection described in the first aspect.

[0033] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for assessing the difficulty of collection described in the first aspect above is implemented.

[0034] The collection difficulty assessment method, device and readable storage medium provided by the present invention first obtain the collection assessment data of overdue customers; then, based on the collection assessment data of the overdue customers, construct the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers; finally, the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customers, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm. The present invention scores the collection difficulty of overdue customers based on the trained collection difficulty model by utilizing the multi-dimensional data of the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of overdue customers, aiming to fully reflect the real collection situation and predict the possibility of future collection of overdue customers, so as to optimize the collection process and improve the collection efficiency, and solve the problem that the existing collection difficulty assessment method has a single assessment dimension, resulting in one-sided assessment results and inability to fully reflect the real collection situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of a method for assessing the difficulty of payment collection in accordance with Embodiment 1 of the present invention;

[0036] Figure 2 A flowchart of another method for assessing the difficulty of payment collection according to an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of the structure of a device for assessing the degree of difficulty of payment collection according to Embodiment 2 of the present invention;

[0038] Figure 4 This is a schematic diagram of the structure of a device for assessing the difficulty of payment collection according to Example 3 of the present invention. DETAILED DESCRIPTION

[0039] In order to enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0040] It should be understood that the specific embodiments and drawings described herein are only used to explain the present invention rather than to limit the present invention.

[0041] It can be understood that, in the absence of conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other.

[0042] It can be understood that, for the convenience of description, the drawings of the present invention only show the parts related to the present invention, while the parts irrelevant to the present invention are not shown in the drawings.

[0043] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0044] It can be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0045] It is understood that the flowcharts and block diagrams of the present invention illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present invention. Each box in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or may be implemented by a combination of hardware and computer instructions.

[0046] It can be understood that the units and modules involved in the embodiments of the present invention can be implemented by software or hardware. For example, the units and modules can be located in a processor.

[0047] Application Overview

[0048] Currently, government and enterprise clients have a large number of accounts receivable generated by overdue users, and they urgently need a scientific and quantified collection strategy to recover the accounts receivable. To this end, we can evaluate the difficulty of collecting payments from government and enterprise clients to predict the possibility of collecting accounts receivable from government and enterprise clients in the future, thereby optimizing the collection process, improving collection efficiency, and improving customer relationships.

[0049] However, the existing technology has the following shortcomings in the application of the collection difficulty model:

[0050] 1. Single evaluation dimension: Existing technologies often evaluate the difficulty of collection based on only a few limited dimensions (such as the number of overdue days, the amount of debt owed, etc.), which leads to one-sided evaluation results and cannot fully reflect the actual collection situation.

[0051] 2. Low data utilization: When processing collection data, existing technologies often fail to fully explore and utilize the potential information in the data, resulting in limited prediction accuracy and reliability of the model.

[0052] 3. Poor adaptability: The existing payment collection difficulty model is usually difficult to adapt to changes in different scenarios and needs, and lacks flexibility and scalability.

[0053] In response to the above technical problems, the present application provides a method, device and readable storage medium for assessing the difficulty of collection. By utilizing multi-dimensional data of overdue customers' overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics, the difficulty of collection for overdue customers is scored based on a trained collection difficulty model. The purpose is to comprehensively reflect the actual collection situation and predict the possibility of future collection from overdue customers, so as to optimize the collection process and improve the collection efficiency, so as to at least solve the problem that the existing collection difficulty assessment method has a single assessment dimension, resulting in one-sided assessment results and failure to fully reflect the actual collection situation.

[0054] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0055] Embodiment 1:

[0056] This embodiment provides a method for assessing the difficulty of payment collection. Figure 1 As shown, the method includes:

[0057] Step S101: Obtain payment collection assessment data of overdue customers.

[0058] It should be noted that the collection assessment data includes: information on arrears and overdue behavior, income and debt information, business cooperation depth information and industrial and commercial information, etc.

[0059] Step S102: Based on the collection assessment data of the overdue customers, construct the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers.

[0060] In this embodiment, by constructing overdue behavior characteristics based on the arrears and overdue behavior information of overdue customers, constructing income and liability characteristics based on the income and liability information of overdue customers, constructing business cooperation characteristics based on the business cooperation depth information of overdue customers, and constructing industrial and commercial characteristics based on the industrial and commercial information of overdue customers, it is possible to fully mine and utilize the collection assessment data and improve data utilization.

[0061] Step S103: input the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm.

[0062] In this embodiment, a collection difficulty model is constructed by adopting a logistic regression algorithm, and the overdue customer's overdue behavior characteristics, income and liability characteristics, business cooperation characteristics, and industrial and commercial characteristics are input into the trained collection difficulty model, so that the trained collection difficulty model scores the overdue customers to obtain a collection difficulty score for the overdue customers, wherein a higher collection difficulty score for the overdue customer represents a greater possibility that the overdue customer will repay.

[0063] In an optional embodiment, before inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes:

[0064] According to the actual business scenario, select the historical collection assessment data of multiple collection customers from the historical data as the modeling sample data;

[0065] According to the modeling sample data, the target variable of the collection difficulty model is determined, and the prediction variables of the collection difficulty model are constructed, wherein the prediction variables include overdue behavior characteristics, income and liability characteristics, business cooperation characteristics, and industrial and commercial characteristics;

[0066] The collection difficulty model is constructed by using a logistic regression algorithm, and the collection difficulty model is trained based on the modeling sample data and the target variable and the predictor variable of the collection difficulty model to obtain the trained collection difficulty model.

[0067] Specifically, historical collection assessment data of multiple collection customers are selected as modeling sample data, where the modeling samples in the modeling sample data will be screened according to different business scenarios. Taking the business scenario of collecting collections from existing customers with serious overdue payments as an example, the number of overdue months and overdue amount of the customers can be conditionally screened: for example, physical customers with a cumulative overdue amount greater than or equal to 500 and a maximum number of overdue months greater than 1 month are selected as modeling samples, and customers with an overdue number of months > 36 months and who have not made any payments in the past year are excluded.

[0068] Specifically, based on the modeling sample data, the repayment behavior and repayment trend of the collection customers corresponding to each modeling sample and the variable distribution of the collection customers corresponding to all modeling samples are defined; and based on the repayment behavior and repayment trend of the collection customers corresponding to each modeling sample and the variable distribution of the collection customers corresponding to all modeling samples, the target variable of the collection difficulty model is determined.

[0069] In this embodiment, the prediction variables of the collection difficulty model are constructed using the modeling sample data, including characteristics in four dimensions: overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics, among which the overdue behavior characteristics include the amount of arrears, the number of overdue days, etc., the income and liability characteristics include the income amount, the amount of bills, etc., the business cooperation characteristics include the number of days of network access, traffic usage, the number of users, etc., and the industrial and commercial characteristics include the registered capital, the registration time, etc. Specifically, the above four dimensional characteristics are calculated as follows:

[0070] 1) Basic statistical features: Calculate the mean, median and other statistics of numerical features, and count and frequency statistics of categorical features;

[0071] 2) Time series characteristics: calculate the frequency of an event within a certain period of time and extract the trend component of the time series;

[0072] 3) Combined features: Perform mathematical operations on multiple numerical features to construct new numerical features;

[0073] 4) Derived features: Based on business logic and common sense, construct derived features that reflect business characteristics.

[0074] Specifically, a logistic regression algorithm is used to construct a collection difficulty model, and based on the modeling sample data and the target variables and predictor variables of the collection difficulty model, model training is implemented through cross-validation to obtain a trained collection difficulty model.

[0075] In an optional embodiment, determining the target variable of the collection difficulty model according to the modeling sample data specifically includes:

[0076] For each collection demand customer in the modeling sample data, the cumulative monthly repayment rate of the collection demand customer is calculated based on the monthly repayment amount and receivable amount in the historical collection demand evaluation data of the collection demand customer.

[0077] Calculate the index average repayment rate of the demanded customers based on the cumulative monthly repayment rate of the demanded customers each month;

[0078] Setting a target threshold based on the distribution of indexed average repayment rates of all the customers called for repayment;

[0079] For each of the collection demanding customers, whether the collection demanding customer is an easy-to-collect customer is determined based on the index average repayment rate of the collection demanding customer and the target threshold, and whether the collection demanding customer is an easy-to-collect customer is used as the target variable.

[0080] Specifically, when determining the target variable of the collection difficulty model, the following factors are mainly considered:

[0081] 1) Repayment behavior: The repayment behavior of customers is measured by defining the cumulative monthly repayment rate of the customers who are being urged to pay each month, that is, the cumulative monthly repayment rate in month T = the repayment amount in month T / the receivable amount in month T;

[0082] Among them, the receivable amount in month T in the historical collection assessment data of the collection customer is the sum of the accumulated arrears amount in month (T-1) and the billing amount in month T in the historical collection assessment data of the collection customer. This indicator reflects the customer's monthly repayment performance. For the historical collection assessment data of each collection customer in the modeling sample data, the cumulative monthly repayment rate of the collection customer is calculated by the above formula every month;

[0083] 2) Repayment trend: We further construct the index average repayment rate of the customers who are being urged to pay (repayment rate trend for short) to measure the ease with which customers are being urged to pay. This is the index average of the cumulative monthly repayment rates of the customers who are being urged to pay each month. It takes into account the continuity of the customer's repayment behavior and can more comprehensively reflect the customer's repayment ability and willingness. For each customer who is being urged to pay in the modeling sample data, the index average repayment rate of the customer who is being urged to pay is calculated using the above formula;

[0084] 3) Variable distribution: According to the distribution of the index average repayment rate of all collection customers, the target threshold is set. For example, assuming that the index average repayment rate of 53% of the collection customers is less than 0.1, and the index average repayment rate of 47% of the collection customers is greater than or equal to 0.1. Then 0.1 is set as the target threshold, so that the balance of the number of good and bad samples in the modeling samples can be maintained. For each collection customer, according to the index average repayment rate of the collection customer and the target threshold, it is judged whether the collection customer is an easy collection customer, and whether the collection customer is an easy collection customer is used as the target variable. Specifically, according to the index average repayment rate of the collection customer and the target threshold, the specific method for judging whether the collection customer is an easy collection customer can be: if the index average repayment rate of the collection customer is greater than or equal to the target threshold, the collection customer is judged to be an easy collection customer; if the index average repayment rate of the collection customer is less than the target threshold, the collection customer is judged to be a difficult collection customer.

[0085] In an optional embodiment, before inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes:

[0086] A scorecard is constructed according to the training results of the trained collection difficulty model, wherein the scoring dimensions of the scorecard include income and liability dimension, business cooperation dimension, overdue behavior dimension and industrial and commercial characteristics dimension.

[0087] It should be noted that the scoring variables of the income and liabilities dimension include the maximum billing amount of physical customers, the total billing amount of natural customers, and the ratio of the maximum amount of overdue payments in the past 12 months to the billing amount in the past 12 months; the scoring variables of the business cooperation dimension include the number of users with normal power on, the total traffic of physical customers, and the number of products; the scoring variables of the overdue behavior dimension include the average number of arrears months, the number of users with one month of arrears, and the minimum overdue amount; the scoring variables of the industrial and commercial characteristics dimension include registered capital and registration duration. Each scoring variable of each scoring dimension of the scorecard has multiple boxes and a numerical range and score weight corresponding to each box.

[0088] In an optional embodiment, the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, specifically including:

[0089] Inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model, and scoring the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer based on the scoring dimensions of the scoring card by the trained collection difficulty model to obtain the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristic score of the overdue customer; and

[0090] The difficulty score of collecting payment from the overdue customer is calculated based on the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristics score of the overdue customer.

[0091] Specifically, the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of overdue customers are input into the trained collection difficulty model. The trained collection difficulty model scores the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of overdue customers according to multiple scoring variables in the income and liability dimensions, business cooperation dimensions, overdue behavior dimensions and industrial and commercial characteristics dimensions of the scoring card, and obtains the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristics score of the overdue customers. Taking the scoring of the industrial and commercial characteristics of overdue customers as an example, the trained collection difficulty model compares the values ​​of the registered capital and registration period in the industrial and commercial characteristics of the overdue customers with the value range corresponding to each bin of the corresponding scoring variable in the industrial and commercial characteristics dimension of the scoring card, and obtains the score weights of the registered capital and registration period in the industrial and commercial characteristics of the overdue customers, and sums the score weights of the registered capital and registration period in the industrial and commercial characteristics of the overdue customers to obtain the industrial and commercial characteristics score of the overdue customers.

[0092] Specifically, the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristic score of the overdue customer are summed up to calculate the difficulty score of collecting payment from the overdue customer.

[0093] It should be noted that after obtaining the difficulty score of collecting payment from overdue customers, the staff can take corresponding measures to collect payment based on the difficulty score of collecting payment from overdue customers and the difficulty of collection corresponding to the difficulty score of collecting payment.

[0094] It is worth mentioning that the collection difficulty assessment method provided by the present invention is based on the collection difficulty model of machine learning logic, which aims to predict the possibility of future collection of customers, so as to optimize the collection process, improve collection efficiency and improve customer relationships. By constructing a collection difficulty model, the collection difficulty of entity customers is predicted by using multi-dimensional data such as overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and operating status characteristics (i.e., industrial and commercial characteristics). The model adopts a logistic regression algorithm, takes whether the entity customer is an easy-to-collect customer as the target variable, constructs relevant prediction variables, and scores each entity customer. The higher the score, the greater the possibility of repayment. The present invention defines the cumulative monthly repayment rate and the target variable (index average repayment rate) to measure the ease of collection of customers, and divides customers into easy-to-collect customers and difficult-to-collect customers accordingly. The present invention constructs a rich feature set through basic statistical features, time series features, combined features and derived features to comprehensively reflect the difficulty of collection of customers.

[0095] In a specific embodiment, a method for assessing the difficulty of collecting accounts receivable of government and enterprise customers is provided. The collection difficulty model based on machine learning logic is intended to predict the possibility of future collection by customers, so as to optimize the collection process and improve the collection efficiency. The present invention predicts the difficulty of collection of entity customers by constructing a collection difficulty model and using multi-dimensional data such as overdue behavior characteristics, income and liability characteristics, business cooperation characteristics, and industrial and commercial characteristics of each customer. The model adopts a logistic regression algorithm, takes whether the entity customer is an easy-to-collect customer as the target variable, constructs relevant predictive variables, and scores each entity customer. The higher the score, the greater the possibility of repayment.

[0096] The original input includes data related to arrears and overdue behavior information, income and debt information, business cooperation depth information, and industrial and commercial information. The final output includes:

[0097] 1) Collection difficulty model, built based on logistic regression algorithm, is used to predict the possibility of future collection of customers.

[0098] 2) Scorecard, which is constructed based on the model training results and is used to quickly assess the difficulty of collecting payments from customers.

[0099] 3) Customer collection difficulty score: Each entity customer is scored, and the higher the score, the greater the possibility of repayment.

[0100] Compared with the existing collection difficulty model, the present invention has a more comprehensive assessment dimension for the collection difficulty of customers, including the income and liability dimension, business cooperation dimension, overdue behavior dimension and industrial and commercial characteristics dimension. The present invention uses a systematic variable derivation method and feature engineering for the original fields of the data, which can fully mine the information of the original fields and greatly improve the utilization efficiency of the data.

[0101] Specifically, Figure 2 As shown in the figure, the specific steps of the collection difficulty assessment method are as follows:

[0102] Step 1: Modeling sample selection and data exploratory analysis

[0103] 1. Modeling sample selection: This model selects customers who need to be focused on collection according to actual business scenarios. Modeling samples will be screened according to different business scenarios. For example, for collection of existing customers with serious overdue payments, the number of overdue months and overdue amount of customers can be conditionally screened: select physical customers with cumulative overdue amount greater than or equal to 500 and the maximum number of overdue months greater than 1 month as modeling samples. Exclude customers with overdue months greater than 36 months and no billing in the past year.

[0104] Among them, the data source: mainly comes from the data middle platform, and historical data is obtained by subscribing to data sets. The data scope and data dictionary: include arrears and overdue behavior information, income and liability information, business cooperation depth information and industrial and commercial information, etc.

[0105] 2. Exploratory Data Analysis (EDA):

[0106] Specifically, explore the data and gain a preliminary understanding of the situation:

[0107] 1) Data check: Confirm the following items: consistency with the data dictionary (table structure, etc.); information adequacy (such as fields, granularity, etc.); data understanding (such as the meaning of the processed fields, statistical caliber, processing logic, etc.); availability (whether there is a need for cleaning work such as deduplication, screening, formatting, missing filling, error and contradiction correction); correlation (by analyzing the correlation between fields / variables, further understand the correlation between data, and consider screening out independent variables with high correlation with the target variable).

[0108] 2) Arrears report: including but not limited to the following dimensions: A. Scale; B. Enterprise type; C. Account age.

[0109] 3) Based on statistical reports, check the overall proportion of arrears, increase and decrease trends, and industry distribution.

[0110] 4) Based on statistical reports, evaluate the recovery rate of overdue payments and the recoverability of overdue payments in specific industries, identify business pain points, and help indicate the focus of the next step of analysis.

[0111] 5) Based on statistical reports, summarize the characteristics of good and bad customers, preliminarily explore important feature directions, and assist in modeling strategy judgment.

[0112] Step 2: Determine the target variable of the collection difficulty model

[0113] Specifically, when constructing the target variable of the collection difficulty model (whether the index average repayment rate is less than 0.1), the following factors are mainly considered:

[0114] 1) Repayment behavior: The customer's repayment behavior is measured by defining the cumulative monthly repayment rate, that is, the cumulative monthly repayment rate in month T = the repayment amount in month T / the receivable amount in month T. Among them, the receivable amount in month T is the sum of the cumulative arrears amount in month (T-1) and the billing amount in month T. This indicator reflects the customer's monthly repayment performance.

[0115] 2) Repayment trend: We further construct an indexed average repayment rate (repayment rate trend for short) to measure the ease with which customers are easily urged to pay. This is the indexed average of the cumulative monthly repayment rates for several consecutive months. It takes into account the continuity of the customer's repayment behavior and can more comprehensively reflect the customer's repayment ability and willingness.

[0116] 3) Variable distribution: According to the distribution of customers’ indexed average repayment rates, 53% of customers have indexed average repayment rates less than 0.1, and 47% of customers have indexed average repayment rates greater than or equal to 0.1. Therefore, using 0.1 as the dividing point to distinguish between difficult / easy customers can maintain the balance of the number of good and bad samples in the modeling samples, which helps the model to more accurately predict the possibility of future collection of customers, thereby optimizing the collection process and improving collection efficiency.

[0117] In summary, the present invention defines customers with an index average repayment rate of ≥ 0.1 as easy-to-collect customers, and customers with an index average repayment rate of < 0.1 as difficult-to-collect customers based on their repayment behavior, repayment trend and variable distribution.

[0118] Step 3: Feature Engineering

[0119] Specifically, construct the predictor variable X: use the data during the observation period to construct features, including features in four dimensions: overdue behavior characteristics (arrears, overdue, etc.), income and liability characteristics (income amount, outgoing amount, etc.), business cooperation characteristics (number of days on the network, traffic, number of users, etc.), industrial and commercial characteristics (registered capital, duration, etc.).

[0120] Specifically, the above four dimensional features are calculated in the following way to obtain more features, such as the average amount of a customer's outstanding payments, the median amount of a customer's outstanding payments, and the amount of a customer's outstanding payments in the past period of time:

[0121] 1) Basic statistical features: Calculate the mean, median and other statistics of numerical features, and count and frequency statistics of categorical features.

[0122] 2) Time series characteristics: calculate the frequency of an event within a certain period of time and extract the trend component of the time series.

[0123] 3) Combined features: Perform mathematical operations on multiple numerical features to construct new numerical features. For example, when constructing the feature "the ratio of the maximum amount of overdue payments in the past 12 months to the amount of bills issued in the past 12 months", the calculation logic of this feature is to take the ratio of the maximum amount of overdue payments of the customer in the past 12 months to the cumulative amount of bills issued in the past 12 months. When constructing this feature, the basic statistics, time series and combination methods of the data are comprehensively considered to form a new hybrid feature, which can fully explore the potential information of the original field and improve the efficiency of data utilization.

[0124] 4) Derived features: Based on business logic and common sense, construct derived features that reflect business characteristics, such as customer overdue payment status, customer traffic usage, etc.

[0125] Specifically, some specific examples of the features in the above four dimensions are shown in Table 1:

[0126] Table 1: Specific examples of some features

[0127]

[0128] Step 4: Model training and scorecard construction

[0129] 1. Model training:

[0130] Specifically, based on the modeling sample of step one, the target variable of the collection difficulty model constructed in step two and the prediction variable of the collection difficulty model constructed in step three, a total of 1099 variables are used in four dimensions: overdue behavior characteristics (arrears, overdue, etc.), income and liability characteristics (income amount, bill amount, etc.), business cooperation characteristics (number of days on the network, traffic, number of users, etc.), and industrial and commercial characteristics (registered capital, duration, etc.). After screening the characteristics through methods such as variance filtering, information value evaluation, and variance inflation test, the target variable of the collection difficulty model (whether the index average repayment rate is less than 0.1) is predicted.

[0131] Specifically, the logistic regression algorithm is used to implement model training through cross-validation: the sample data during model training is divided into a training set and a test set. A common method is to randomly divide the data set into a training set and a test set. For example, 70% of the data can be used as a training set and the remaining 30% as a test set. The advantage of this method is that it is simple and easy to implement, and it can ensure the statistical independence of the training set and the test set.

[0132] Finally, 11 variables in 4 dimensions were selected as modeling variables. A scorecard was constructed based on the regression coefficients and evidence weights of the modeling variables. An example of the scorecard is shown in Table 2:

[0133] Table 2: Scorecard Example

[0134]

[0135]

[0136] Step 5: Output and Application:

[0137] Specifically, taking the entity customers of Company A as an example, the collection difficulty model proposed in the present invention is used to predict the collection difficulty. The specific method is as follows:

[0138] 1. Data collection and preprocessing: Subscribe to relevant data sets from the data center, including arrears details, billing data, user information, industrial and commercial information, etc., and perform data cleaning and preprocessing.

[0139] 2. Feature construction: Based on the collected data, basic statistical features, time series features, combined features and derived features are constructed.

[0140] 3. Model training and scoring output: Use the logistic regression algorithm to train the constructed features, obtain the collection difficulty model, and score each entity customer.

[0141] 4. Application of results: Apply the model scoring results to the scenario of targeted marketing of accounts receivable collection. The account manager can take appropriate measures to collect the collection according to the collection difficulty score and collection difficulty. An example of customer collection difficulty score is shown in Table 3:

[0142] Table 3: Example of customer collection difficulty rating

[0143]

[0144] The collection difficulty assessment method provided by the embodiment of the present invention first obtains the collection assessment data of overdue customers; then, based on the collection assessment data of the overdue customers, constructs the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers; finally, the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customers, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm. The present invention scores the collection difficulty of overdue customers based on the trained collection difficulty model by utilizing the multi-dimensional data of the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of overdue customers, aiming to comprehensively reflect the real collection situation and predict the possibility of future collection of overdue customers, so as to optimize the collection process and improve the collection efficiency, and solves the problem that the existing collection difficulty assessment method has a single assessment dimension, resulting in one-sided assessment results and failure to fully reflect the real collection situation.

[0145] Embodiment 2:

[0146] like Figure 3 As shown, this embodiment provides a device for assessing the difficulty of payment collection, which is used to execute the above-mentioned method for assessing the difficulty of payment collection, including:

[0147] An acquisition module 11 is used to acquire payment collection assessment data of overdue customers;

[0148] A construction module 12, connected to the acquisition module 11, is used to construct the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers according to the collection assessment data of the overdue customers;

[0149] The evaluation module 13 is connected to the construction module 12, and is used to input the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers into the trained collection difficulty model to obtain the collection difficulty score of the overdue customers, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm.

[0150] Furthermore, the device also includes:

[0151] The selection module is used to select historical collection assessment data of multiple collection customers from historical data as modeling sample data according to actual business scenarios;

[0152] A determination module, used to determine the target variable of the collection difficulty model according to the modeling sample data, and construct the prediction variables of the collection difficulty model, wherein the prediction variables include overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics;

[0153] The training module is used to construct the collection difficulty model by using a logistic regression algorithm, and train the collection difficulty model according to the modeling sample data and the target variable and the predictor variable of the collection difficulty model to obtain the trained collection difficulty model.

[0154] Furthermore, the determination module specifically includes:

[0155] A first calculation unit is used to calculate the cumulative monthly repayment rate of each collection customer in the modeling sample data according to the repayment amount and receivable amount of each month in the historical collection assessment data of each collection customer;

[0156] A second calculation unit is used to calculate the index average repayment rate of the urged customer based on the cumulative monthly repayment rate of the urged customer each month;

[0157] A setting unit, used for setting a target threshold value according to the distribution of the index average repayment rate of all the collection customers;

[0158] The judging unit is used to judge, for each of the collection customers, whether the collection customer is an easy-to-collect customer according to the index average repayment rate of the collection customer and the target threshold, and use whether the collection customer is an easy-to-collect customer as the target variable.

[0159] Further, the determination as a unit includes:

[0160] A first judgment unit is used to judge that the collection customer is an easy collection customer if the index average repayment rate of the collection customer is greater than or equal to the target threshold;

[0161] The second judgment unit is used to judge that the demanding customer is a difficult-to-demand customer if the index average repayment rate of the demanding customer is less than the target threshold.

[0162] Furthermore, the device also includes:

[0163] A scoring card module is constructed to construct a scoring card according to the training results of the trained collection difficulty model, wherein the scoring dimensions of the scoring card include income and liability dimensions, business cooperation dimensions, overdue behavior dimensions and industrial and commercial characteristics dimensions.

[0164] Furthermore, the evaluation module 13 specifically includes:

[0165] a scoring calculation module, for inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model, and scoring the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer based on the scoring dimensions of the scoring card through the trained collection difficulty model to obtain the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristic score of the overdue customer; and,

[0166] The difficulty score of collecting payment from the overdue customer is calculated based on the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristics score of the overdue customer.

[0167] Furthermore, the device also includes:

[0168] The collection module is used to take corresponding measures to collect the overdue customers according to the collection difficulty score.

[0169] Embodiment 3:

[0170] refer to Figure 4 This embodiment provides a device for assessing the difficulty of payment collection, including a memory 21 and a processor 22. The memory 21 stores a computer program, and the processor 22 is configured to run the computer program to execute the method for assessing the difficulty of payment collection in Example 1.

[0171] The memory 21 is connected to the processor 22. The memory 21 may be a flash memory, a read-only memory or other memory. The processor 22 may be a central processing unit or a single-chip microcomputer.

[0172] Embodiment 4:

[0173] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for assessing the difficulty of collection in the above-mentioned embodiment 1 is implemented.

[0174] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0175] In summary, the collection difficulty assessment method, device and readable storage medium provided in the embodiment of the present invention first obtain the collection assessment data of overdue customers; then, based on the collection assessment data of the overdue customers, construct the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers; finally, the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customers, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm. The present invention scores the collection difficulty of overdue customers based on the trained collection difficulty model by utilizing the multi-dimensional data of the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of overdue customers, aiming to fully reflect the actual collection situation and predict the possibility of future collection of overdue customers, so as to optimize the collection process and improve the collection efficiency, and solve the problem that the existing collection difficulty assessment method has a single assessment dimension, resulting in one-sided assessment results and inability to fully reflect the actual collection situation.

[0176] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for assessing the difficulty of payment collection, characterized in that: The method comprises: Obtain collection assessment data for overdue customers; Based on the collection assessment data of the overdue customers, construct the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers; The overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, wherein the collection difficulty model is a model built based on a logistic regression algorithm.

2. The method according to claim 1, characterized in that: Before inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes: According to the actual business scenario, select the historical collection assessment data of multiple collection customers from the historical data as the modeling sample data; According to the modeling sample data, the target variable of the collection difficulty model is determined, and the prediction variables of the collection difficulty model are constructed, wherein the prediction variables include overdue behavior characteristics, income and liability characteristics, business cooperation characteristics, and industrial and commercial characteristics; The collection difficulty model is constructed by using a logistic regression algorithm, and the collection difficulty model is trained based on the modeling sample data and the target variable and the predictor variable of the collection difficulty model to obtain the trained collection difficulty model.

3. The method according to claim 2, characterized in that Determining the target variable of the collection difficulty model according to the modeling sample data specifically includes: For each collection demand customer in the modeling sample data, the cumulative monthly repayment rate of the collection demand customer is calculated based on the monthly repayment amount and receivable amount in the historical collection demand evaluation data of the collection demand customer. Calculate the index average repayment rate of the demanded customers based on the cumulative monthly repayment rate of the demanded customers each month; Setting a target threshold based on the distribution of indexed average repayment rates of all the customers called for repayment; For each of the collection demanding customers, whether the collection demanding customer is an easy-to-collect customer is determined based on the index average repayment rate of the collection demanding customer and the target threshold, and whether the collection demanding customer is an easy-to-collect customer is used as the target variable.

4. The method according to claim 3, characterized in that The determining, based on the index average repayment rate of the demanded customer and the target threshold, whether the demanded customer is an easy-to-demand customer specifically includes: If the index average repayment rate of the collection customer is greater than or equal to the target threshold, the collection customer is judged to be an easy-to-collect customer; If the index average repayment rate of the collection customer is less than the target threshold, the collection customer is determined to be a difficult-to-collect customer.

5. The method according to claim 1, characterized in that Before inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes: A scorecard is constructed according to the training results of the trained collection difficulty model, wherein the scoring dimensions of the scorecard include income and liability dimension, business cooperation dimension, overdue behavior dimension and industrial and commercial characteristics dimension.

6. The method according to claim 5, characterized in that The overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer are input into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, specifically including: Inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model, and scoring the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer based on the scoring dimensions of the scoring card by the trained collection difficulty model to obtain the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristic score of the overdue customer; and The difficulty score of collecting payment from the overdue customer is calculated based on the overdue behavior score, income and liability score, business cooperation score and industrial and commercial characteristics score of the overdue customer.

7. The method according to claim 1, characterized in that After inputting the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customer into the trained collection difficulty model to obtain the collection difficulty score of the overdue customer, the method further includes: Based on the difficulty level of collection for overdue customers, corresponding measures will be taken to collect the payment.

8. A device for assessing the difficulty of payment collection, characterized in that: include: The acquisition module is used to obtain the collection assessment data of overdue customers; A construction module, connected to the acquisition module, for constructing overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers based on the collection assessment data of the overdue customers; The evaluation module is connected to the construction module and is used to input the overdue behavior characteristics, income and liability characteristics, business cooperation characteristics and industrial and commercial characteristics of the overdue customers into the trained collection difficulty model to obtain the collection difficulty score of the overdue customers, wherein the collection difficulty model is a model constructed based on a logistic regression algorithm.

9. A device for assessing the difficulty of payment collection, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the method for assessing the difficulty of collection as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for assessing the difficulty of collection according to any one of claims 1 to 7 is implemented.

Images