Service-configuration-based bill payment system

Abstract

Provided in the present application is a service-configuration-based bill payment system, which is configured to: on the basis of service configuration parameters of a bill payment system, acquiring a transaction risk level and a user credit score; by means of a dynamic decision-making rule, determining whether the transaction risk level and the user credit score trigger an insurance mechanism; using a machine learning algorithm to train a transaction problem probability prediction model on the basis of historical transaction data; for current batch-processed bills, using the transaction problem probability prediction model to predict a transaction problem occurrence probability; if the transaction problem occurrence probability exceeds a preset threshold value, triggering the insurance mechanism, otherwise, dynamically adjusting a batch processing quantity on the basis of a bill amount and the user credit score; and by means of an intelligent routing algorithm, comprehensively taking the transaction risk level, the user credit score and a payment guarantee level into account to dynamically route the bills.

Description

A bill payment system based on business configuration TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a bill payment system based on business configuration. BACKGROUND

[0002] There is a technical contradiction between pursuing batch processing efficiency and providing insurance guarantee in a bill payment system. Batch processing can significantly improve the transaction processing capacity and efficiency of the system by packaging multiple transactions together for processing. However, batch processing also means higher transaction risk. Once a problem occurs in a transaction in the batch, it may cause a chain reaction leading to the failure of the entire batch, causing greater losses to the system and even users. The original intention of the insurance mechanism is to reduce the transaction risk of users, but providing insurance will inevitably consume system resources and reduce system performance. TECHNICAL PROBLEM

[0003] How the system balances batch processing and insurance mechanism is a problem that needs to be deeply thought according to the business scenarios of the system. Bill payment systems often face numerous merchants and a large number of C-end users, with different transaction modes and risk levels. For transactions with low risk levels and good user credit, the system can reduce the priority of insurance triggering and use batch processing to pursue efficiency as much as possible. For high-risk transactions and users with poor credit records, the priority of insurance triggering should be increased to control risks using the insurance mechanism. This requires the system to be able to control in multiple dimensions such as transaction risk level and user credit score for different business scenarios, dynamically decide whether to use batch processing and whether to trigger the insurance mechanism, so as to achieve a dynamic balance between batch efficiency and risk prevention and control, and meet differentiated business needs. TECHNICAL SOLUTION

[0004] The present application provides a bill payment system based on business configuration, mainly comprising:

[0005] obtaining a transaction risk level and a user credit score according to bill payment system business configuration parameters;

[0006] determining whether the insurance mechanism is triggered by the transaction risk level and the user credit score through a dynamic decision rule;

[0007] training a transaction problem probability prediction model based on historical transaction data using a machine learning algorithm;

[0008] predicting the probability of transaction problems occurring using the transaction problem probability prediction model for the current batch of bills;

[0009] if the transaction problem probability exceeds a preset threshold to trigger the insurance mechanism, otherwise the batch processing quantity is dynamically adjusted according to the bill amount and the user credit score;

[0010] The bill is dynamically routed by an intelligent routing algorithm considering the transaction risk level, the user credit score and the payment guarantee degree. Advantages

[0011] The technical scheme provided by the embodiment of the application can include the following advantages:

[0012] The application discloses an intelligent bill processing method based on risk assessment. The method first acquires a transaction risk level and a user credit score, and judges whether to trigger an insurance mechanism through a dynamic decision rule. At the same time, a machine learning algorithm is used to train a transaction problem probability prediction model to predict the current batch of bills. If the prediction probability exceeds a threshold, the insurance mechanism is triggered, otherwise the batch processing quantity is dynamically adjusted according to the bill amount and the credit score. Finally, the risk level, the credit score and the payment guarantee degree are comprehensively considered, and an intelligent routing algorithm is used to dynamically route the bills. The application realizes the automation, intelligence and risk controllability of bill processing through multi-dimensional risk assessment and intelligent decision-making, effectively improves the bill processing efficiency and security, and provides users with a better payment experience. BRIEF DESCRIPTION OF DRAWINGS

[0013] Fig. 1 is a flowchart of a bill payment system based on business configuration according to the application.

[0014] Fig. 2 is a schematic diagram of a bill payment system based on business configuration according to the application.

[0015] Fig. 3 is another schematic diagram of a bill payment system based on business configuration according to the application. Embodiments of the application

[0016] For a further understanding of the present application, reference will be made to the following detailed description of the application taken in conjunction with the accompanying drawings. The following detailed description of the application is provided for the purpose of illustration only and will not be interpreted to limit the scope of the application. It should be appreciated that the specific examples described herein are intended to be illustrative only and are not intended to limit the scope of the application. In addition, it should be noted that, for the purpose of clarity, only the parts of the drawings that are necessary for the understanding of the application are shown.

[0017] As shown in Figs. 1-3, the bill payment system based on business configuration according to the embodiment can specifically include:

[0018] S101, acquiring a transaction risk level and a user credit score according to bill payment system business configuration parameters.

[0019] According to the transaction request, the transaction amount, the transaction time and the transaction object information are obtained, historical transaction data and user credit records are combined, a logistic regression model is used to evaluate the transaction risk, and a transaction risk score is obtained; according to the transaction risk level, different measures are taken: if the transaction risk score is greater than or equal to a preset first risk threshold and less than a preset second risk threshold, the user is required to perform identity verification; if the transaction risk score is greater than or equal to the preset second risk threshold, the upper limit of the transaction amount is limited; the transaction risk score obtained by evaluation, together with the transaction amount and the transaction object information, is used as the input of the logistic regression model, a random forest algorithm is used to calculate the credit score of the user in real time, and the credit score of the user is dynamically updated according to the calculation result; the transaction risk score and the user credit score are included in the business rule engine of the bill payment, and the system strategy including the transaction limit and the transaction permission is dynamically adjusted according to the preset rule; after the transaction is completed, the transaction result and the user feedback are obtained, the transaction result and the user feedback are used as sample data, the logistic regression model is periodically retrained, and the model performance is continuously optimized; the business rules and threshold settings of the bill payment system are periodically audited, the matching degree of the risk rules and the business requirements is sorted out, and when the matching degree of the risk rules and the business requirements does not meet the requirements, the rules are adjusted to improve the risk prevention and control ability of the system.

[0020] Specifically, in the bill payment system, when receiving a shopping transaction request of 500 yuan, the system first obtains the amount, time and identity information of the transaction object of the transaction. Then, the system retrieves the user's transaction records in the past 6 months from the historical transaction database and finds that the user has made 3 similar shopping transactions with amounts of 480 yuan, 550 yuan and 520 yuan, respectively, and has completed the payment on time. At the same time, the system also learns from the user's credit file that the user's credit score is 750 points (total score 800 points), belonging to the creditworthy user group. Based on the above information, the system uses a logistic regression model to assess the risk of the transaction, and after calculation, the risk score of the transaction is 35, which is greater than or equal to the first risk threshold 30 and less than the second risk threshold 40, so it is determined that the transaction risk is low. However, the system still requires the user to perform SMS verification for identity authentication to confirm the user's identity. If the risk score of the transaction is calculated to be 45, which is greater than or equal to the second risk threshold 40, the system limits the upper limit of the transaction amount; after identity authentication, the system uses a random forest algorithm to calculate the user's credit score in real time using 10 features such as the user's transaction frequency, transaction amount, credit score, etc. as input, and obtains the user's latest credit score of 770 points. According to the preset business rules, the system dynamically adjusts the user's single transaction limit to 800 yuan, and successfully releases the 500 yuan transaction request. After the transaction is completed, the system also obtains the user's 5-star good review of this shopping experience, which is included in the model training set as sample data. Through periodic model retraining, the system optimizes the performance of the logistic regression risk assessment model and the random forest credit score model, ensuring business continuity while improving the system's risk prevention and control capabilities.

[0021] S102、Through a dynamic decision rule, determine whether the transaction risk level and the user credit score trigger an insurance mechanism.

[0022] Obtain transaction-related data in a transaction system, including transaction amount, transaction time, transaction object, transaction type and transaction location;

[0023] Preprocess the transaction-related data, including data cleaning, feature extraction and normalization processing of the transaction-related data;

[0024] Input the preprocessed transaction-related data into a preset transaction risk assessment model, wherein the transaction risk assessment model uses a random forest algorithm, and outputs a transaction risk level corresponding to the transaction-related data;

[0025] obtain user historical transaction data and credit records in a user information system, the user historical transaction data and credit records including user past transaction frequency, transaction amount, transaction type, and credit default situation;

[0026] input the user historical transaction data and credit records into a preset credit scoring model, the credit scoring model adopting a logistic regression algorithm, and output a credit score corresponding to the user;

[0027] perform weighted calculation on the transaction risk level and the credit score according to a preset risk control rule to obtain a comprehensive risk index;

[0028] if the comprehensive risk index exceeds a preset threshold, trigger a corresponding insurance mechanism, the insurance mechanism including requiring additional identity verification, limiting transaction amount, and delaying transaction execution.

[0029] Specifically, in a bill payment system of an e-commerce platform, when a 1000-yuan shopping transaction request is received, the system first obtains that the transaction amount is 1000 yuan, the transaction time is May 10, 2023, 14:30, the transaction object is a well-known brand clothing store, the transaction type is clothing consumption, and the transaction location is Haidian District, Beijing. Then, the system preprocesses these transaction-related data, removes outliers in the transaction amount and transaction time through data cleaning, extracts five key features such as transaction amount and transaction type, and normalizes the feature data by using the maximum and minimum value normalization method. Then, the system inputs the preprocessed transaction data into a transaction risk assessment model based on a random forest algorithm, which outputs the risk level of the transaction as “low risk” by comprehensively analyzing the features such as transaction amount and transaction type. At the same time, the system obtains from the user information system that the user’s transaction frequency in the past year is 5 times per month, the monthly transaction amount is 3000 yuan, the main transaction type is clothing consumption, and there is no credit default record. The system inputs these user historical transaction data and credit records into a credit scoring model based on a logistic regression algorithm, and obtains a credit score of 750 points (total score 800 points) for the user. Next, the system performs weighted calculation on the transaction risk level “low risk” and the user credit score 750 points according to the preset risk control rule, and obtains a comprehensive risk index of 20, which does not exceed the preset risk threshold 50, so it is determined that the transaction is a normal transaction, and the payment process is successfully released and completed. Through the bill payment system, the e-commerce platform can monitor the risk status of each transaction in real time, and dynamically adjust the risk control strategy according to the credit situation of the user, improving the payment efficiency while effectively preventing fraud risks.

[0030] S103, training a transaction problem probability prediction model based on historical transaction data using a machine learning algorithm.

[0031] Historical transaction data is collected and preprocessed. Preprocessing includes deleting duplicate records, filling missing values, and standardizing numerical features. Key features such as transaction amount, transaction frequency, and transaction time are extracted from the preprocessed data. The extracted features are combined with known problem transaction labels to form a training dataset. Based on the characteristics of the training dataset, a random forest algorithm is selected to build a transaction problem probability prediction model. Random forest can handle high-dimensional features, is not sensitive to outliers, and has generalization ability. The training data is input into the random forest algorithm, and the initial model is trained by adjusting the number of trees and the depth of the trees. The model performance is evaluated using five-fold cross-validation. The accuracy, precision, recall, and F1 value of each validation set are calculated. Based on the evaluation results, the model parameters such as the number of trees and the maximum depth are optimized using grid search to improve the prediction performance of the model. For new transaction data, the same preprocessing and feature extraction methods as the training data are applied. The extracted features are input into the optimized transaction problem probability prediction model to obtain the probability value of the transaction having a problem. Set the risk probability threshold to 0.7. If the predicted probability is greater than the threshold, it is determined as a high-risk transaction. For high-risk transactions, the system automatically sends risk prompt information to relevant operating personnel and suspends the subsequent processing flow of the transaction. New transaction data and problem feedback are collected daily. When the cumulative amount of new data reaches 10% of the original training set, the model is updated. During the update, the new data is combined with the original training set, and the model training and optimization process is repeated. If the F1 value of the updated model improves by more than 1%, the existing model is replaced; otherwise, the original model is retained. The model prediction results are manually audited regularly (e.g., monthly). Based on the audit feedback, adjust the feature extraction method or risk probability threshold. If a new fraud pattern is found, it will be included in the feature engineering process to improve the model's ability to identify new problem transactions.

[0032] Specifically, in the transaction risk monitoring system of an e-commerce platform, 5 million historical transaction records from the past year were first retrieved from the transaction database. This data was then preprocessed, with data cleaning algorithms removing 2% of duplicate records and 5% of missing values. Numerical features such as transaction amount and time were normalized to their maximum and minimum values. Next, 10 key features, including transaction amount, transaction frequency, transaction time, and transaction object, were extracted from the preprocessed data and combined with corresponding labels indicating whether a transaction had a problem, forming a training dataset containing 4.5 million records. Considering the high dimensionality of this dataset and the potential for outliers, a random forest algorithm was chosen to construct the transaction problem prediction model. By setting parameters such as the number of trees to 100 and the maximum depth to 20, the model was trained on the training dataset, resulting in an initial random forest model. To evaluate the model's performance, a five-fold cross-validation method was used, randomly dividing the training set into five subsets. Four subsets were used to train the model each time, and the remaining subset was used for validation. After five cross-validations, the model achieved an average accuracy of 95%, precision of 91%, recall of 89%, and an F1 score of 90% on the validation set. To improve model performance, a grid search algorithm was used to optimize parameters such as the number of trees and the maximum depth, resulting in a random forest model with 150 trees and a maximum depth of 25. When new transaction data arrives, the system automatically performs the same preprocessing and feature extraction as the training data, then inputs the extracted features into the optimized random forest model for prediction. If the probability of a transaction being predicted as a problematic transaction is greater than 70%, the system classifies it as a high-risk transaction, automatically sends risk warnings to relevant personnel, and suspends the payment process for that transaction. Furthermore, the system automatically collects newly generated transaction data and user-reported problematic transaction data daily. When the accumulated new data reaches 10% of the original training set, i.e., 450,000 records, the system automatically triggers a model update process, merging the new data into the original training set for retraining and optimization. After a model update, the updated model's F1 score on the validation set improved from 90% to 92%, an improvement of over 1%. Therefore, the system automatically replaced the updated model with the new online prediction model. Simultaneously, the system also extracts 1000 transaction data points predicted as high-risk each month for manual review. Based on the review results, the risk assessment threshold and feature extraction methods are dynamically adjusted to continuously improve the model's risk identification capabilities.

[0033] S104. For the current batch processing bill, use the transaction problem probability prediction model to predict the probability of a transaction problem occurring.

[0034] For batch processing of billing transactions, first obtain the attributes of each transaction from the transaction database, including transaction object, transaction time, transaction amount, and recent transaction frequency. Clean and standardize the obtained transaction attribute data, including removing outliers, filling missing values, and unifying data formats. Input the processed transaction attribute data into the pre-trained transaction problem probability prediction model, which is trained based on historical transaction data to predict the probability of each transaction problem. Set the problem occurrence probability threshold to 0.7, and determine whether the predicted probability of each transaction exceeds the threshold. For transactions exceeding the threshold, mark them as high-risk transactions and obtain additional features, including the credit score of the transaction object, the difference between the transaction amount and the average transaction amount of the account in the past 3 months, whether the transaction time is in an irregular period, etc. Input all features of high-risk transactions (including initial attributes and additional features) into the pre-trained gradient boosting decision tree model, which is used to calculate the suspiciousness score of high-risk transactions. Sort the high-risk transactions in descending order according to the suspiciousness score, and select the top 10% transactions with the highest scores as suspicious transactions that need to be focused on and verified. For the selected suspicious transactions, generate an early warning report containing transaction details and suspiciousness score. Through the application programming interface of the enterprise internal instant messaging system, push the early warning report in the form of a message to the responsible person of the relevant business department. At the same time, create a verification task in the pending task list of the transaction monitoring system and assign it to a specific business personnel. After receiving the early warning information, the business personnel logs in to the transaction monitoring system and views the detailed information of the suspicious transaction. According to the pre-set verification process, the business personnel needs to contact the transaction related parties, verify the transaction background, and check relevant documents to verify the authenticity and compliance of the transaction. The verification result will be recorded in the transaction monitoring system, including transaction status (normal, abnormal, fraud, etc.), verification process description, and processing suggestions. For transactions confirmed as abnormal or fraudulent, the system automatically generates a processing work order and transfers it to the risk control department for investigation and processing. At the same time, feedback the relevant transaction information to the model training module to update and optimize the transaction problem probability prediction model regularly, improve the accuracy and adaptability of the model.

[0035] Specifically, in the transaction risk monitoring system of a certain cross-border e-commerce platform, first, through the data interface with the transaction system, the attribute data of each cross-border transaction is obtained in real time, including the registered country of the transaction object, the transaction time, the transaction amount (in US dollars), and the transaction frequency of the account in the last week. Then, the natural language processing technology is used to standardize the registered country name of the transaction object, unify different languages, synonyms, and abbreviations into standardized English country names; the transaction time is converted to Greenwich Mean Time; the transaction amount is converted to US dollars; and the transaction frequency is detected for outliers, eliminating extreme values exceeding 3 times the standard deviation of the average value. Next, the processed transaction attribute data is input into a random forest model of 100 trees trained based on 100 million historical transaction data, which has an AUC (Area Under Curve) of 0.95 on the test set, used to predict the probability of each transaction problem. If the probability of a transaction problem exceeds the threshold of 0.7, it is marked as a high-risk transaction, and through the query interface of the transaction system, additional features such as the credit score of the transaction object (provided by a third-party credit investigation agency, ranging from 0 to 100), the difference between the transaction amount and the average daily transaction amount of the account in the last 3 months, and whether the transaction time is in the high-risk period from 23:00 to 5:00 the next day are obtained. Then, the 7 features of the high-risk transaction (4 initial attributes + 3 additional features) are input into a gradient boosting decision tree model consisting of 500 trees and a maximum tree depth of 8, which has an accuracy of 92% on the validation set, used to calculate the suspiciousness score of the high-risk transaction (ranging from 0 to 100). According to the suspiciousness score, all high-risk transactions on the same day are sorted in descending order, and the top 10% (if less than 100, all are selected, and at most 500) are selected as suspicious transactions, generating an early warning report containing information such as transaction serial number, transaction object account number, transaction amount, transaction time, and suspiciousness score. Through the RestFul API interface of the internal IM system, the early warning report is pushed to the responsible person of the cross-border business department in the form of a rich text message in Json format, and is sent with a CSV format attachment. At the same time, the early warning report information is automatically written into the pending task table of the Mysql database in the transaction monitoring system, and the trigger is automatically assigned to a specific business personnel, with the status set to "to be checked". After receiving the early warning message through the IM system, the business personnel can log in to the Web page of the transaction monitoring system by clicking the link in the message, find the corresponding early warning transaction in the pending task list, and view the transaction details. The page is embedded with data interfaces of the enterprise's CRM system, ERP system, logistics system, etc., and the business personnel can retrieve the historical transaction data, commodity inventory status, logistics order number, etc. of the transaction object with one click, or call the transaction object's phone or send an email, etc., to efficiently complete the verification.After the verification is completed, the business personnel selects the transaction status (normal / abnormal / fraud) on the page, inputs the verification result description and processing suggestions, and clicks submit. If abnormal or fraud is selected, the system automatically creates a risk event record and pushes it to the risk control department for processing. At the same time, the actual status and related information of the transaction are transmitted to the model training module in real time through the Kafka real-time data stream in the form of Json. After receiving the feedback data, the model training module starts the feature engineering process, redesigns the features and merges them into the model training set. When the cumulative feedback data reaches 100,000, the model retraining is automatically started, the gradient boosting decision tree model is updated, and the updated model parameters are written into the configuration file, which takes effect in real time.

[0036] S105, if the transaction problem occurrence probability exceeds the preset threshold, triggering the insurance mechanism, otherwise dynamically adjusting the batch processing quantity according to the bill amount and the user credit score.

[0037] The transaction problem occurrence probability is obtained, which is compared with the preset threshold. If the transaction problem occurrence probability exceeds the preset threshold, the insurance mechanism is triggered, and the corresponding risk prevention measures are executed. If the transaction problem occurrence probability does not exceed the preset threshold, the subsequent transaction processing process is continued. The bill amount and user credit score data of the current transaction are obtained as input features, and a pre-trained CART decision tree model is used for prediction to obtain the batch processing quantity adjustment amplitude for the current transaction. The batch processing quantity represents the number of transaction records contained in each batch when the transaction data is processed in batches. The batch processing quantity adjustment amplitude predicted by the decision tree model is weighted averaged with the currently set batch processing quantity, and the weight is set according to the historical data to obtain a new batch processing quantity. The adjusted batch processing quantity is applied to the transaction processing process, and the corresponding batch processing operation is performed on the transaction data, improving the transaction processing efficiency. The change of the transaction problem occurrence probability is continuously monitored. When the occurrence probability increases, the grid search algorithm is used to optimize the preset threshold and the hyperparameters of the CART decision tree model, and the threshold and the adjustment strategy of the batch processing quantity are dynamically updated to adapt to the changes of the transaction mode, ensuring the dynamic balance between risk prevention and transaction efficiency.

[0038] Specifically, in the transaction risk monitoring system of a certain cross-border e-commerce platform, first, through the data interface with the transaction system, the attribute data of each cross-border transaction is obtained in real time, including the registered country of the transaction object, the transaction time, the transaction amount (in US dollars) and the transaction frequency of the account in the last week. Then, the natural language processing technology is used to standardize the registered country name of the transaction object, unify different languages, synonyms and abbreviations into standard English country names; the transaction time is converted to Greenwich Mean Time; the transaction amount is converted to US dollars; the transaction frequency is detected for outliers, and extreme values exceeding 3 times the standard deviation of the average value are removed. Next, the processed transaction attribute data is input into a random forest model of 100 trees trained based on 10 million historical transaction data, the AUC of the model on the test set reaches 0.95, which is used to predict the probability of each transaction problem. If the probability of a transaction problem exceeds the preset threshold of 0.8, the insurance mechanism is triggered, and the relevant information of the transaction is automatically sent to the insurance company to apply for starting the corresponding risk compensation process; if it does not exceed the threshold, the subsequent transaction processing is continued. At the same time, the system obtains the billing amount and user credit score of the current transaction as input features, uses a CART decision tree model trained based on 50,000 historical data to predict the adjustment amplitude of the batch processing quantity for the current transaction. For example, if the current batch processing quantity is set to 50 batches, the decision tree model predicts an adjustment amplitude of +10%, then the original batch quantity is weighted and averaged according to a weight of 7:3 to obtain a new batch processing quantity of 52 batches. The system continuously monitors the average transaction problem probability every hour, and when the probability increases by more than 15% compared with the previous hour, a grid search algorithm is automatically started to perform 10-fold cross-validation in the parameter range of threshold 0.7-0.9, decision tree maximum depth 3-8, minimum leaf node sample size 10-100, etc. to obtain the optimal model hyperparameter combination, and real-time update the preset threshold and batch quantity adjustment strategy to dynamically adapt to the transaction mode changes.

[0039] S106, dynamically routing the bill by comprehensively considering the transaction risk level, the user credit score, and the payment guarantee degree through an intelligent routing algorithm.

[0040] Get the information of the bill to be processed, including transaction amount, transaction type, user historical behavior, etc. Extract the user's transaction records in the past half year from the database, calculate the transaction frequency, average transaction amount, etc. According to the transaction frequency and average transaction amount features, use the logistic regression model to calculate the transaction risk score. Compare the risk score with the preset threshold to determine the transaction risk level. Get the user's credit score from the credit scoring system. The credit scoring system calculates a score of 0-100 based on the user's repayment history and credit card usage. Determine the payment security level of the transaction. Check if the transaction supports refund and whether it is guaranteed. Assign a value of 1 to transactions that support refund and a value of 0 to transactions that do not support refund. Assign a value of 1 to transactions that are guaranteed and a value of 0 to transactions that are not guaranteed. Add the value of whether the transaction supports refund and whether it is guaranteed to get the payment security degree, ranging from 0 to 2. Take the transaction risk level, user credit score, and payment security degree as input variables. Use the random forest algorithm to build a decision model. The output of the decision model is the probability distribution of different processing channels. Select the channel with the highest probability as the optimal routing scheme. According to the routing scheme, allocate the bill to the corresponding processing channel. The specific implementation is as follows: Establish a processing channel mapping table to correspond the channel ID output by the decision model with the API interface of the actual processing system. Call the corresponding API to transmit the bill information to the target processing system. After the processing system receives the bill, return the processing result. Record the processing result to the database for subsequent model optimization and system evaluation. Analyze the processing result regularly to evaluate the accuracy and efficiency of the routing scheme and provide basis for model parameter adjustment.

[0041] Specifically, in the intelligent billing routing system of a certain cross-border e-commerce platform, first, through the data interface with the transaction system, the attribute information of each pending bill is obtained in real time, including the transaction amount, transaction type (such as purchasing goods, recharging, etc.) and user ID. Then, the system extracts the user's transaction records in the past half year from the database, aggregates and counts them by month using SQL statements, and obtains the user's monthly transaction frequency and monthly transaction amount, etc. characteristic variables. Next, these characteristic variables are input into a logistic regression model trained based on 100,000 historical data, which has an accuracy rate of 85% on the test set, used to predict the risk score of each transaction, with a score range of 0 to 100. The system sets three risk levels: low risk (0-30), medium risk (30-70), and high risk (70-100). At the same time, the system obtains the user's credit score from the credit scoring system through the API interface, with a score range of 350 to 950 points. In addition, the system will also judge the payment security level of the transaction, assigning a value of 1 to transactions that support refunds and 0 to those that do not; a value of 1 to transactions with third-party guarantees and 0 to those without, and adding the two values to obtain the payment security degree. Finally, the system uses a random forest model with 100 trees trained based on 50,000 historical data, taking the transaction risk level, user credit score, and payment security degree as input variables, to predict the probability distribution of different processing channels, and selects the channel with the highest probability as the optimal routing solution. For example, for a low-risk transaction with a transaction amount of $500, a user credit score of 650 points, and a payment security degree of 1, the model calculates that the probability of being assigned to the manual review channel is 2%, the probability of being automatically passed is 95%, and the probability of being rejected is 3%, so the transaction will be automatically passed. According to the routing result, the system calls the corresponding API interface, transmits the bill information in JSON format to the target processing system, and waits for the return processing result. The system automatically analyzes the bill processing results of the previous hour every 1 hour, calculates the actual pass rate, rejection rate, etc. of each channel, and if the deviation from the expected value exceeds 5%, the model parameters are automatically adjusted to adapt to changes in bill distribution.

[0042] Obviously, those skilled in the art can make various modifications and variations to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include these modifications and variations.

Claims

1. A service configuration based billing payment system, characterized by, The bill payment system comprises: According to the bill payment system business configuration parameter, the transaction risk level and the user credit score are obtained; Determine whether the transaction risk level and the user credit score trigger the insurance mechanism through dynamic decision rules; Adopting machine learning algorithm, training transaction problem probability prediction model based on historical transaction data; For the current batch processing bill, the transaction problem probability prediction model is used to predict the probability of transaction problem; If the transaction problem probability exceeds the preset threshold, the insurance mechanism is triggered, otherwise the batch processing quantity is dynamically adjusted according to the bill amount and the user credit score; Through intelligent routing algorithm, the transaction risk level, the user credit score and the payment guarantee degree are comprehensively considered to dynamically route the bill; Wherein, it also includes: obtaining the transaction problem probability, comparing it with the preset threshold, if the transaction problem probability exceeds the preset threshold, triggering the insurance mechanism, and executing the corresponding risk prevention measures; If the transaction problem probability does not exceed the preset threshold, the subsequent transaction processing flow is continued; Obtain the bill amount and user credit score data of the current transaction as input features, and use the pre-trained CART decision tree model for prediction to obtain the batch processing quantity adjustment range for the current transaction; The batch processing quantity represents the number of transaction records contained in each batch when the transaction data is processed in batches; The batch processing quantity adjustment range predicted by the decision tree model is weighted averaged with the current batch processing quantity, and the weight is set according to the historical data to obtain the new batch processing quantity; Apply the adjusted batch processing quantity to the transaction processing flow to perform corresponding batch processing operation on the transaction data, and improve the transaction processing efficiency; Continuously monitor the change of the transaction problem probability, when the probability increases, use grid search algorithm to optimize the preset threshold and the hyperparameters of CART decision tree model, dynamically update the threshold and the adjustment strategy of batch processing quantity, to adapt to the change of transaction mode, and ensure the dynamic balance of risk prevention and transaction efficiency.

2. The bill payment system of claim 1, wherein, According to the bill payment system business configuration parameter, the transaction risk level and the user credit score are obtained, including: According to the transaction request, the transaction amount, transaction time and transaction object information are obtained, combined with the historical transaction data and user credit record, the transaction risk is evaluated using the logistic regression model to obtain the transaction risk score; According to the transaction risk level, different measures are taken: if the transaction risk score is greater than or equal to the preset first risk threshold and less than the preset second risk threshold, the user is required to perform identity verification; If the transaction risk score is greater than or equal to the preset second risk threshold, the upper limit of the transaction amount is limited; The transaction risk score obtained by evaluation, together with the transaction amount and transaction object information, are used as the input of the logistic regression model, and the user's credit score is calculated in real time using the random forest algorithm, and the user's credit score is dynamically updated according to the calculation result; The transaction risk score and the user credit score are included in the bill payment business rule engine, and the system strategy including the transaction limit and the transaction right is dynamically adjusted according to the preset rule; After the transaction is completed, the transaction result and user feedback are obtained, the transaction result and the user feedback are taken as sample data, the logistic regression model is retrained regularly, and the model performance is continuously optimized; Periodically review the business rules and threshold settings of the bill payment system, and analyze the matching degree of risk rules and business needs. When the matching degree of risk rules and business needs does not meet the requirements, adjust the rules to improve the risk prevention and control capability of the system.

3. The bill payment system of claim 1, wherein, The method comprises the following steps: Obtain transaction-related data in the transaction system, wherein the transaction-related data comprises transaction amount, transaction time, transaction object, transaction type and transaction location; Preprocess the transaction-related data, wherein the preprocessing comprises data cleaning, feature extraction and normalization processing of the transaction-related data; Input the preprocessed transaction-related data into a preset transaction risk assessment model, wherein the transaction risk assessment model adopts a random forest algorithm, and outputs a transaction risk level corresponding to the transaction-related data; Obtain user historical transaction data and credit records in the user information system, wherein the user historical transaction data and credit records comprise user past transaction frequency, transaction amount, transaction type and credit default situation; Input the user historical transaction data and credit records into a preset credit scoring model, wherein the credit scoring model adopts a logistic regression algorithm, and outputs a credit score corresponding to the user; According to a preset risk control rule, the transaction risk level and the credit score are weighted calculated to obtain a comprehensive risk index; If the comprehensive risk index exceeds a preset threshold, a corresponding insurance mechanism is triggered, and the insurance mechanism comprises requiring additional identity verification, limiting transaction amount and delaying transaction execution.

4. The bill payment system of claim 1, wherein, The method comprises the following steps: Obtain historical transaction data and preprocess the data; Preprocessing includes deleting duplicate records, filling missing values and standardizing numerical features; Extract key features from the preprocessed data, including transaction amount, transaction frequency and transaction time; Combine the extracted key features with known problem transaction labels to form a training data set; Based on the characteristics of the training data set, select a random forest algorithm to build a transaction problem probability prediction model; Random forest algorithm can handle high-dimensional features, is not sensitive to outliers, and has generalization ability; Input the training data into the random forest algorithm, and train an initial model by adjusting the number of trees and the depth of the trees; Use five-fold cross-validation method to evaluate the model performance; Calculate the accuracy, precision, recall and F1 value on each validation set; According to the evaluation results, use grid search method to optimize the model parameters, including the number of trees and the maximum depth, to improve the prediction performance of the model; For new transaction data, use the same preprocessing and feature extraction method as the training data; Input the extracted features into the optimized transaction problem probability prediction model to obtain the probability value of the transaction problem. Set the risk probability threshold to 0.7; If the predicted probability is greater than the threshold, it is determined to be a high-risk transaction; For high-risk transactions, the system automatically sends risk warning information to relevant personnel and suspends the subsequent processing of the transaction; Collect new transaction data and problem feedback daily; When the cumulative amount of new data reaches 10% of the original training set, trigger model update; When updating, combine new data with the original training set, and repeat the model training and optimization process; If the F1 value of the updated model improves by more than 1%, replace the existing model; Otherwise, keep the original model; Periodically review the model's prediction results manually; Adjust the feature extraction method or risk probability threshold based on the audit feedback; If a new fraud pattern is found, it will be included in the feature engineering process to improve the model's ability to identify new problem transactions.

5. The bill payment system of claim 1, wherein, The transaction problem probability prediction model is used to predict the probability of a transaction problem for the current batch of processed bills, including: For batch-processed bill transactions, first obtain the attributes of each transaction from the transaction database, including transaction object, transaction time, transaction amount, and recent transaction frequency; Clean and standardize the obtained transaction attribute data, including removing outliers, filling missing values, and unifying data formats; Input the processed transaction attribute data into the pre-trained transaction problem probability prediction model, which is trained based on historical transaction data to predict the probability of each transaction problem; Set the problem occurrence probability threshold to 0.7 to determine whether the predicted probability of each transaction exceeds the threshold; For transactions that exceed the threshold, mark them as high-risk transactions and obtain additional features, including the credit score of the transaction object, the difference between the transaction amount and the average transaction amount of the account in the past three months, and whether the transaction time is in an irregular period; Input all features of high-risk transactions into the pre-trained gradient boosting decision tree model, which is used to calculate the suspiciousness score of high-risk transactions; Sort high-risk transactions in descending order based on suspiciousness scores, and select the top 10% transactions with the highest scores as suspicious transactions that need to be closely monitored and verified; For the selected suspicious transactions, generate a warning report containing transaction details and suspiciousness scores; Through the application programming interface of the enterprise's instant messaging system, push the warning report to the responsible person of the relevant business department in the form of a message; At the same time, create a verification task in the pending task list of the transaction monitoring system and assign it to a specific business personnel; After receiving the warning information, the business personnel logs in to the transaction monitoring system to view the detailed information of the suspicious transaction; According to the pre-set verification process, the business personnel needs to verify the authenticity and compliance of the transaction by contacting the relevant parties, verifying the transaction background, and checking relevant documents; The verification results will be recorded in the transaction monitoring system, including transaction status, verification process description, and processing suggestions; For transactions confirmed as abnormal or fraudulent, the system automatically generates a processing work order and transfers it to the risk control department for investigation and processing; At the same time, feedback the relevant transaction information to the model training module for regular updating and optimization of the transaction problem probability prediction model to improve the accuracy and adaptability of the model.

6. The bill payment system of claim 1, wherein, The dynamic routing of the bill by the intelligent routing algorithm considering the transaction risk level, the user credit score, and the payment guarantee degree includes: Obtaining the to-be-processed bill information, including the transaction amount, the transaction type, and the user historical behavior; Extracting the transaction records of the user in the past half year from the database, calculating the transaction frequency and the average transaction amount characteristics; According to the transaction frequency and the average transaction amount characteristics, using a logistic regression model to calculate the transaction risk score; Comparing the risk score with a preset threshold to determine the transaction risk level; Obtaining the credit score of the user from the credit scoring system; The credit scoring system calculates a score of 0-100 based on the user's repayment history and credit card usage using the weighted average method; Judging the payment guarantee degree of the transaction; Checking whether the transaction supports refund and whether it is guaranteed; Assigning a value of 1 to the transaction that supports refund and a value of 0 to the transaction that does not support refund; Assigning a value of 1 to the transaction that is guaranteed and a value of 0 to the transaction that is not guaranteed; Adding the value of whether the transaction supports refund and the value of whether it is guaranteed to obtain the payment guarantee degree, which ranges from 0 to 2; Taking the transaction risk level, the user credit score, and the payment guarantee degree as input variables; Using a random forest algorithm to build a decision model; The output of the decision model is the probability distribution of different processing channels; Selecting the channel with the highest probability as the optimal routing scheme; According to the routing scheme, the bill is allocated to the corresponding processing channel; The specific implementation is to establish a processing channel mapping table, which corresponds the channel ID output by the decision model to the API interface of the actual processing system; Calling the corresponding API to transmit the bill information to the target processing system; After the processing system receives the bill, the processing result is returned; Recording the processing result to the database for subsequent model optimization and system evaluation; Periodically analyzing the processing result to evaluate the accuracy and efficiency of the routing scheme and provide a basis for model parameter adjustment.

Images