Coin counting prepayment management method and system

Abstract

This invention discloses a digital currency prepayment management method and system. The method includes the following steps: S1, acquiring multi-dimensional real-time data in a digital currency prepayment scenario, and preprocessing the data by cleaning, normalizing, and format standardization; S2, inputting the preprocessed multi-dimensional data into a trained machine learning risk prediction model, and using the model to extract features and calculate risk probabilities; S3, determining the risk level based on the risk probabilities, and generating early warning prompts or automatically triggering corresponding control measures according to preset risk response rules. This invention relates to the field of digital currency prepayment management technology. This digital currency prepayment management method and system, through comprehensive collection and cleaning of multi-dimensional real-time data, combined with risk prediction using a machine learning model, can efficiently and accurately identify potential risks, avoiding the misjudgments and omissions that may occur in traditional early warning mechanisms based on simple rules, and improving the accuracy of risk prediction.

Description

Technical Field

[0001] This invention relates to the field of digital currency prepayment management, and more specifically, to a digital currency prepayment management method and system. Background Technology

[0002] With the widespread adoption of digital currencies, digital prepayment, as an emerging payment method, is increasingly used in numerous consumption scenarios. However, digital prepayment scenarios face many risks, potentially causing losses for both users and financial institutions.

[0003] At the user level, there may be situations such as abnormal consumption behavior and threats to account fund security. For example, users may fall victim to fraud, resulting in the illegal transfer of prepaid funds; or their own consumption habits may suddenly change, resulting in abnormally large prepaid transactions that they fail to notice in time.

[0004] From the perspective of financial institutions and regulators, the lack of effective risk monitoring and control measures makes it difficult to grasp the true risk status of digital currency prepayment transactions in real time. Traditional risk assessment methods are often based on limited, single-dimensional data, which cannot comprehensively and accurately reflect transaction risks. For example, judging solely based on account balance or transaction amount easily overlooks other important factors, such as user consumption behavior patterns and social network connections, leading to inaccurate and untimely risk warnings and an inability to effectively prevent potential risks.

[0005] Furthermore, existing risk response measures are relatively simplistic, lacking flexibility and specificity. For transactions of different risk levels, differentiated control strategies cannot be adopted based on the actual situation; they are either too lenient, failing to effectively curb risk, or too strict, affecting normal trading and reducing user experience.

[0006] Therefore, there is an urgent need for a digital currency prepayment management method and system that can integrate multi-dimensional data, accurately predict risks, reasonably determine risk levels, and take effective countermeasures to ensure the safe and stable operation of digital currency prepayment transactions. Summary of the Invention

[0007] The purpose of this invention is to provide a digital currency prepayment management method and system, which solves the problems of existing risk response measures being relatively simple, lacking flexibility and specificity. They usually adopt a single rule or a model prediction based on past transaction experience, which is difficult to provide comprehensive and accurate risk prediction in complex real-time data environments and cannot meet the needs of users.

[0008] This invention achieves the above objective through the following technical solution: a digital currency prepayment management method, the method comprising the following steps: S1. Obtain multi-dimensional real-time data in the digital currency prepayment scenario, and perform preprocessing such data cleaning, normalization and format standardization. S2. Input the preprocessed multi-dimensional data into the trained machine learning risk prediction model, and use the model to extract features and calculate risk probabilities from the data. S3. Determine the risk level based on the risk probability, and generate early warning prompts or automatically trigger corresponding control measures according to preset risk response rules; S4. Real-time feedback of processing results to the user end and system management end, and record risk processing logs for subsequent traceability and model optimization.

[0009] Furthermore, the multi-dimensional real-time data includes: User consumption behavior data, account balance data, transaction frequency data, cryptocurrency market data, and user social network association data; Among them, user consumption behavior data includes consumption amount, consumption scenario, consumption time distribution, and transaction counterparty information; Account balance data includes real-time balance, balance change trend, and prepaid funds ratio, wherein the prepaid funds ratio is the ratio of the amount of prepaid funds locked in the account to the total account balance; The trading frequency data includes the number of transactions per unit time and the trading interval fluctuation coefficient, which reflects the dispersion of each trading interval within a unit time. The digital currency market data includes real-time exchange rates, price volatility, and market liquidity indicators. The price volatility reflects the fluctuation range of digital currency prices within a unit of time. User social network association data includes transaction frequency of associated accounts, percentage of common trading counterparties, and abnormal social relationship markers. The percentage of common trading counterparties is the ratio of the number of common trading counterparties between the target user and associated accounts to the total number of trading counterparties of the target user.

[0010] Furthermore, the data cleaning includes: Remove missing values, outliers, and duplicate data; use interpolation to fill in missing data for key dimensions; and mark outliers that exceed a reasonable range. The normalization process is to convert data with different dimensions to a unified numerical range to eliminate the impact of differences in dimensions. The format standardization includes converting data from different sources into a preset timestamp format, monetary unit, and account identifier representation to ensure data format consistency.

[0011] Furthermore, the training process of the machine learning risk prediction model includes: Collect historical digital currency prepayment transaction data and corresponding risk event markers, divide the training set and test set, and ensure that the proportion of risk events is consistent; A hybrid model architecture combining gradient boosting trees and neural networks is adopted. The gradient boosting trees determine the feature importance weights, the neural networks output the preliminary risk probability, and the results of the two are combined to calculate the final risk probability. A specific loss function is used for model training, and cross-validation is used to avoid overfitting and adjust hyperparameters. The accuracy, recall, and F1 score of the model are calculated using the test set to verify the model's performance.

[0012] Furthermore, the feature importance weights are determined by the information gain ratio, and the weights, after normalization, sum to 1. The final risk probability is obtained by weighted summation of the risk probability output by the gradient boosting tree and the initial risk probability output by the neural network. The weighting coefficients are determined to be optimal through cross-validation combined with grid search. The hyperparameters include learning rate, tree depth, and number of hidden layer nodes in the neural network.

[0013] Furthermore, the step of determining the risk level based on risk probability includes: Clustering algorithms are used to cluster the probability distribution of historical risk events, and the initial risk level threshold is determined after obtaining the cluster centers. The initial risk levels include at least four categories: low risk, medium risk, high risk, and extremely high risk. The threshold is dynamically adjusted each preset period based on the handling effect of risk events in the previous month to ensure that the threshold is within a preset reasonable range and meets the level differentiation logic.

[0014] Furthermore, the preset risk response rules include: When the risk probability is within the low-risk threshold range, only the risk assessment result is recorded, and no proactive warning is triggered; When the risk probability is within the medium risk threshold range, a risk warning message is sent to the user to remind them to verify the transaction behavior, and a risk warning notification is pushed to the management end at the same time. When the risk probability is within the high-risk threshold range, the upper limit of the single prepayment amount is limited. The upper limit is dynamically determined based on the current available account balance and the risk probability range, and is not lower than the preset minimum prepayment amount. Warning information is also sent to the user and the management terminal simultaneously. When the risk probability exceeds the extremely high risk threshold, the digital currency prepayment function will be frozen immediately, new prepayment transactions will be prohibited, a freeze notification and risk explanation will be sent to the user, and a manual review process will be triggered on the management side. The freeze can only be lifted after the review is approved.

[0015] Furthermore, the methods for providing real-time feedback processing results include: Feedback is provided to users through multiple communication channels, including APP push, SMS, and in-app messages, with feedback delays controlled within a preset time. Provide the management with a visual interface for risk management, displaying risk events, handling measures, and user feedback in real time.

[0016] Furthermore, the risk handling log shall include at least the following contents: Risk assessment time, multi-dimensional data details, risk probability, risk level, handling measures, execution results, and user operation feedback; Logs can be filtered and retrieved by time range, account ID, risk level, etc., and the retrieval response time does not exceed the preset time. Log data is stored in a distributed manner to ensure security, and the data retention period is no less than the preset number of years.

[0017] A digital currency prepayment management system, applied to perform the above-described digital currency prepayment management method, the system comprising: Data acquisition unit, data preprocessing unit, risk prediction unit, risk level determination unit, risk response unit, and feedback recording unit; The data acquisition unit is used to acquire multi-dimensional real-time data in the digital currency prepayment scenario; The data preprocessing unit is used to clean, normalize, and standardize the data. The risk prediction unit is used to perform feature extraction and risk probability calculation through a trained machine learning risk prediction model. The risk level determination unit is used to determine the risk level based on the risk probability; The risk response unit is used to generate early warning prompts or trigger control measures according to preset rules; The feedback recording unit is used to provide real-time feedback of processing results to the user end and the management end, and to record risk processing logs.

[0018] The beneficial effects of this invention are as follows: 1. By comprehensively collecting and cleaning multi-dimensional real-time data and combining it with machine learning models for risk prediction, potential risks can be identified efficiently and accurately. This avoids the misjudgments and omissions that may occur in traditional early warning mechanisms based on simple rules, thus improving the accuracy of risk prediction.

[0019] 2. It can provide real-time feedback on risk handling results, and promptly provide feedback to users through various communication channels such as APP push, SMS and in-site messages, ensuring timely response from users and the management side, and displaying risk events and handling measures in real time, which helps the management side to monitor and optimize decisions in real time.

[0020] 3. Based on different risk levels, the system takes different measures through preset response rules, such as limiting the amount of a single prepayment, freezing the digital currency prepayment function, and triggering a manual review process, to ensure the security of digital currency prepayment transactions. Especially in the face of high-risk events, it can quickly take control measures to reduce potential losses.

[0021] 4. Record detailed risk handling logs, including risk assessment time, data details, risk probability, risk level, and handling measures, which provides reliable data support for subsequent traceability and model optimization and helps to continuously improve the accuracy of risk prediction models.

[0022] 5. By using a hybrid model of gradient boosting trees and neural networks, and combining it with cross-validation to avoid overfitting, the model can be trained and optimized using historical data. The model can be continuously adjusted based on new data and events, thereby continuously improving prediction performance.

[0023] 6. By performing cluster analysis on the probability distribution of historical risk events and dynamically adjusting the risk level thresholds, we can ensure that the distinction between risk levels is more accurate and reasonable, thereby improving the flexibility and adaptability of risk management. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the data preprocessing sub-process of the present invention; Figure 2 This is a diagram of the risk model training architecture of the present invention; Figure 3 This is a diagram illustrating the risk level assessment and response for this invention. Figure 4 This is a system block diagram of the present invention. Detailed Implementation

[0025] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0026] Example 1: Please see Figure 1-3 This invention provides a technical solution: a digital currency prepayment management method, the method comprising: S1. Obtain multi-dimensional real-time data in digital currency prepayment scenarios, and perform data cleaning, normalization, and format standardization preprocessing. Among them, digital currency prepayment scenarios refer to business scenarios related to using digital currency for prepayment; multi-dimensional real-time data is data collected from multiple different aspects and angles that can reflect the status of digital currency prepayment scenarios in real time. This data may include information from different dimensions such as transaction amount, transaction time, transaction location, identity information of the transacting parties, and transaction frequency; data cleaning involves checking, filtering, and correcting the collected raw data to remove inaccurate or incomplete data such as noise, erroneous data, and duplicate data, thereby improving data quality; normalization is the transformation of data according to certain rules to make data of different dimensions comparable, usually by mapping data to a specific range to avoid the impact of different data dimensions on subsequent analysis; format standardization is the uniform conversion of data into a specific format, such as a unified date format or a unified data type, to facilitate data storage, processing, and analysis. S2. Input the preprocessed multi-dimensional data into the trained machine learning risk prediction model, and use the model to extract features and calculate risk probabilities from the data. The trained machine learning risk prediction model is developed by training and optimizing a machine learning algorithm using a large amount of digital currency prepayment scenario data with known risk labels. This allows the algorithm to learn the intrinsic relationship between features and risks in the data, thus constructing a model capable of predicting risks for new data. Feature extraction involves automatically identifying and extracting key information features that play a crucial role in risk prediction from preprocessed multi-dimensional data using the machine learning model. These features can more effectively reflect the inherent patterns of the data and their correlation with risks. Risk probability calculation is based on the extracted features. The machine learning risk prediction model calculates the probability value of a risk occurring in the digital currency prepayment transaction or related business based on the patterns and rules it has learned internally. S3. Determine the risk level based on the risk probability, and generate early warning prompts or automatically trigger corresponding control measures according to preset risk response rules; The risk level is determined by calculating the risk probability and classifying risks into different levels according to pre-set criteria. Different risk levels correspond to different degrees of risk and scope of impact. Pre-defined risk response rules are pre-established strategies and operational procedures for different risk levels. These rules specify the concrete measures to be taken when different risk situations occur; for example, low-risk situations may only require a warning, while high-risk situations may directly restrict transactions. Early warning alerts are notifications sent to relevant personnel when the calculated risk probability reaches a certain threshold, triggering pre-defined warning conditions and reminding them of potential risks. Automatic triggering of corresponding control measures involves the system automatically executing corresponding control and management measures based on the determined risk level and pre-defined risk response rules. These measures may include freezing accounts, limiting transaction amounts, and requiring additional verification to reduce the likelihood of risk occurrence or minimize potential losses. S4. Real-time feedback of processing results to the user end and system management end, and record risk processing logs for subsequent traceability and model optimization; The real-time feedback processing results are promptly communicated to users and system administrators regarding the warnings or control measures taken to address risks, allowing relevant personnel to understand the risk handling situation in a timely manner. The risk handling log records in detail relevant information for each risk identification, warning, and control process in the digital currency prepayment scenario, including the time of risk occurrence, risk type, risk level, measures taken, and handling results. This log data can be used for subsequent traceability queries to understand the entire risk handling process, and also provides data support for optimizing machine learning risk prediction models. By analyzing historical risk handling data, shortcomings of the model can be identified and improved.

[0027] It should be noted that when using this technology, acquiring and preprocessing multi-dimensional real-time data ensures data quality, provides a reliable foundation for subsequent analysis, and makes risk assessment more accurate. Utilizing a trained machine learning risk prediction model for feature extraction and risk probability calculation can automatically uncover potential patterns in the data, efficiently and accurately identify risks, reduce the subjectivity and error of human judgment, determine the risk level based on the risk probability, and generate early warnings or trigger control measures. This enables timely response to risks, reduces losses, ensures the safety and stability of digital currency prepayment services, and provides real-time feedback on processing results and log recording. This allows users and management to understand the situation promptly, enhances trust and control capabilities, and provides a basis for subsequent traceability and model optimization, continuously improving the risk management system, enhancing the overall risk prevention and control level, and promoting the healthy development of digital currency prepayment services.

[0028] In one embodiment, acquiring multi-dimensional real-time data in a digital currency prepayment scenario includes: The data collected should at least cover user consumption behavior data, account balance data, transaction frequency data, cryptocurrency market data, and user social network association data. Among them, user consumption behavior data includes consumption amount, consumption scenario, consumption time distribution, and transaction counterparty information; Account balance data includes real-time balance, balance change trend, and prepaid funds ratio. The formula for calculating the prepaid funds ratio is as follows: in, The percentage of prepaid funds, This refers to the amount of prepaid funds already locked in the account. Total account balance; Trading frequency data includes the number of trades per unit time and the trading interval volatility coefficient. The formula for calculating the trading interval volatility coefficient is as follows: in, This is the volatility coefficient for the trading interval. The standard deviation of each transaction interval within a unit of time. This represents the average of all transaction intervals within a unit of time. Cryptocurrency market data includes real-time exchange rates, price volatility, and market liquidity indicators. The formula for calculating price volatility is: , in, For price volatility, The highest price of a currency within a unit of time. The lowest price of a currency within a unit of time. The average price of a number of currencies per unit of time; User social network association data includes transaction frequency of associated accounts, percentage of common trading counterparties, and abnormal social relationship markers. The formula for calculating the percentage of common trading counterparties is as follows: in, The percentage of common counterparties, The number of common trading counterparties between the target user and associated accounts. The total number of counterparties for the target user.

[0029] This design collects data from multiple dimensions, covering user consumption, accounts, transaction frequency, market conditions, and social network connections. Each dimension of data is defined in detail. This multi-dimensional data can comprehensively reflect the status of digital currency prepayment scenarios. User consumption behavior data can provide insights into consumption habits, account and transaction frequency data reflects fund flow and activity, market condition data is linked to external factors, and social network connection data can uncover potential risk associations. By combining these data, a rich basis is provided for subsequent accurate risk assessment, enabling more sensitive detection of potential risks, improving the accuracy and comprehensiveness of risk identification, and ensuring the safe and stable operation of digital currency prepayment business.

[0030] In one embodiment, the data is preprocessed by cleaning, normalizing, and format standardization, including: Data cleaning: Remove missing values, outliers, and duplicate data. For data missing in key dimensions, use linear interpolation to fill in the gaps. The interpolation formula is: in, For the first Interpolation results for each missing location, For the first A known data value, For the first A known data value, ; Outlier data exceeding 3 standard deviations are marked. The formula for outlier identification is as follows: in, Data to be determined. This represents the mean of the data in this dimension. The standard deviation of the data in this dimension; Normalization process: For data of different dimensions, the min-max normalization method is used to transform them to the [0,1] interval. The normalization expression is as follows: in, The original data, The maximum value of the data in this dimension. The minimum value of the data in this dimension. The data is after normalization; Standardized format: Data from different sources is uniformly converted into a preset data format, ensuring that the timestamp format is YYYY-MM-DD-HH:MM:SS, the monetary unit is uniformly the target digital currency unit, and the account identifier is represented by a cryptographic hash value.

[0031] This design involves data cleaning to remove bad data and using linear interpolation to fill in missing values ​​and mark outliers; normalization to convert data of different dimensions to the [0,1] interval; and format standardization to unify the data format. Data cleaning improves data quality, ensures the reliability of analysis, and avoids interference from bad data. Normalization eliminates the influence of dimensions, making different features comparable and facilitating model training. Format standardization unifies the data format, facilitates data storage, processing, and sharing, and improves system compatibility and data processing efficiency. The combination of these three aspects provides a high-quality and standardized data foundation for subsequent model training and risk assessment, thereby improving overall management effectiveness.

[0032] In one embodiment, the training process of the trained machine learning risk prediction model includes: Building the training dataset: Collect historical digital currency prepayment transaction data and corresponding risk event markers, divide the training set and test set in a 7:3 ratio, and use stratified sampling to ensure that the proportion of risk events in the training set and test set is consistent; Model selection and training: A hybrid model architecture combining gradient boosting trees (XGBoost) and neural network MLPs is adopted, with the gradient boosting tree outputting feature importance weights. , For the feature dimension, the feature importance weight is determined by the information gain ratio, and the calculation formula is: in, For the first Information gain of each feature For the first The information entropy of each feature, after weight normalization, satisfies... ; The neural network outputs a preliminary risk probability. The final risk probability of the hybrid model is calculated as follows: in, To increase the risk probability of the tree output by gradient enhancement, The weight system has a range of values. The optimal value is determined by combining 5-fold cross-validation with grid search, with a search step size of 0.05, and the goal is to maximize the F1 score on the validation set. The model training loss function uses the binary cross-entropy loss function: in, The number of training samples. For the first The true label of each sample (1 for risk events and 0 for non-risk events). The probability of risk predicted by the model; Model validation and optimization: Calculate the model's accuracy using the test set. Recall rate and F1 score The calculation formulas are as follows: in, This represents the number of true positive samples. This represents the number of true negative samples. This represents the number of false positive samples. This represents the number of false negative samples. Five-fold cross-validation is used to avoid overfitting, and hyperparameters such as learning rate (0.001-0.1), tree depth (3-10), and number of hidden layer nodes in the neural network (64-256) are adjusted by grid search.

[0033] This design involves constructing a training dataset, selecting a hybrid model architecture, determining weight coefficients, training with a binary cross-entropy loss function, determining hyperparameters through model validation and optimization, constructing a training set through stratified sampling to ensure data representativeness, and combining the advantages of gradient boosting trees and neural networks to enhance risk prediction capabilities. The determined weight coefficients allow the model to integrate the outputs of both, resulting in more accurate calculation of risk probabilities. The binary cross-entropy loss function is suitable for binary classification problems and facilitates model convergence. Five-fold cross-validation and grid search optimize hyperparameters, avoiding overfitting and improving model generalization ability, thereby more accurately predicting digital currency prepayment risks and providing strong support for risk management.

[0034] In one embodiment, determining the risk level based on risk probability includes: Initial determination rules for risk level thresholds: The K-means clustering algorithm was used to cluster the probability distribution of historical risk events (K=4) to obtain the cluster centers. Sort by ascending order, and take the median of adjacent cluster centers as the threshold, i.e.: The initial threshold range is limited to , , If the clustering result exceeds the range, the boundary value is taken; The default initial threshold is set to , , ; Threshold dynamic adjustment rules: The threshold is updated monthly based on the effectiveness of risk event handling in the previous month, defining the handling accuracy rate. ,when When adjusting, use the following formula: The adjusted threshold must meet the following requirements. If the value exceeds the range, it will be truncated to the boundary value; When the model outputs the risk probability At that time, it was determined to be low risk; when At that time, it was determined to be a medium-risk area; when At that time, it was determined to be high-risk; when At that time, it was determined to be extremely high risk.

[0035] This design first determines the initial threshold, then dynamically adjusts it each calendar month based on the handling effect. The risk level is determined according to the relationship between the risk probability and the threshold. The initial threshold is determined by a clustering algorithm combined with historical data distribution, which has a certain degree of rationality. The dynamic adjustment rule updates the threshold based on the handling accuracy rate, making the risk level classification more in line with the actual situation, improving the accuracy and adaptability of risk assessment, reflecting risk changes in a timely manner, providing a basis for taking targeted measures for different levels of risk, effectively preventing the expansion of risks, and ensuring the stable operation of digital currency prepayment business in a dynamic environment.

[0036] In one embodiment, generating an early warning prompt or automatically triggering corresponding control measures based on preset risk response rules includes: Low risk, specifically Only risk assessment results are recorded; no proactive warnings are triggered. Medium risk, specifically : Send risk warning messages to users, reminding them to verify their transaction behavior, and push risk warning notifications to the management side; High risk, specifically The maximum amount that can be prepaid in a single transaction is limited. The formula for calculating the maximum amount is as follows: in, This is the maximum amount that can be prepaid in a single transaction. The current available account balance is used. When the calculated result is less than the preset minimum prepayment amount, the default value is 100 times the minimum unit of the target currency, and the minimum prepayment amount is taken. A warning message is also sent to the user and the management terminal simultaneously. Extremely high risk, specifically Immediately freeze the digital currency prepayment function, prohibit the initiation of new prepayment transactions, send a freeze notification and risk explanation to users, and trigger a manual review process on the management side. The freeze can only be lifted after the review is approved.

[0037] This design sets different response measures according to different risk levels, gradually escalating the control intensity from recording results to freezing functions. The tiered response measures are reasonable and targeted: low-risk cases are only recorded to reduce interference with users; medium-risk cases alert users and management to raise awareness; high-risk cases restrict prepayment amounts to reduce the risk of loss; and extremely high-risk cases freeze functions and require manual review to ensure security. This tiered approach can effectively prevent risks while balancing user experience and business operation needs, minimizing the impact on normal business while ensuring security, and improving the risk management level of digital currency prepayment business.

[0038] In one embodiment, the processing results are fed back to the user terminal and the system management terminal in real time, and a risk processing log is recorded, including: Feedback method: The processing results will be fed back to users through three channels: APP push, SMS and in-app message, with a feedback delay of no more than 3 seconds; Provide a risk management dashboard to the management team, displaying risk events, handling measures, and user feedback in real time; Log records: The recorded information includes risk assessment time, multi-dimensional data details, risk probability, risk level, handling measures, execution results, user operation feedback, etc. Log retrieval supports filtering by time range, account ID, and risk level, with a retrieval response time of no more than 1 second; Log data is retained for no less than 3 years, and distributed storage is used to ensure data security.

[0039] This design allows for rapid feedback of processing results to users through multiple channels, provides dashboards for management, records detailed logs with support for quick retrieval and long-term storage, and enables users and management to stay informed about risk handling status in a timely manner, enhancing trust and control capabilities. The management dashboard displays information intuitively, facilitating real-time monitoring and decision-making. Detailed log records provide rich data for subsequent traceability, analysis, and model optimization, while the quick retrieval function improves work efficiency. Long-term storage ensures data integrity. The overall design enhances the transparency, efficiency, and traceability of risk management, promoting the continued healthy development of digital currency prepayment business.

[0040] Example 2: Please see Figure 4 A digital currency prepayment management system, characterized in that it is applied to the above-described digital currency prepayment management method, and the system includes: Data acquisition unit, data preprocessing unit, risk prediction unit, risk level determination unit, risk response unit, and feedback recording unit; The data acquisition unit is used to acquire multi-dimensional real-time data in the digital currency prepayment scenario; The data preprocessing unit is used to clean, normalize, and standardize the format of the data. The risk prediction unit is used to extract features and calculate risk probabilities using a trained machine learning risk prediction model. Risk level determination unit, used to determine the risk level based on risk probability; The risk response unit is used to generate early warning prompts or trigger control measures based on preset rules. The feedback recording unit is used to provide real-time feedback on the processing results to the user end and the management end, and to record the risk processing log.

[0041] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0042] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for managing digital currency prepayments, characterized in that, The method includes the following steps: S1. Obtain multi-dimensional real-time data in the digital currency prepayment scenario, and perform preprocessing such data cleaning, normalization and format standardization. S2. Input the preprocessed multi-dimensional data into the trained machine learning risk prediction model, and use the model to extract features and calculate risk probabilities from the data. S3. Determine the risk level based on the risk probability, and generate early warning prompts or automatically trigger corresponding control measures according to preset risk response rules; S4. Real-time feedback of processing results to the user end and system management end, and record risk processing logs for subsequent traceability and model optimization.

2. The digital currency prepayment management method according to claim 1, characterized in that, The multi-dimensional real-time data includes: User consumption behavior data, account balance data, transaction frequency data, cryptocurrency market data, and user social network association data; Among them, user consumption behavior data includes consumption amount, consumption scenario, consumption time distribution, and transaction counterparty information; Account balance data includes real-time balance, balance change trend, and prepaid funds ratio, wherein the prepaid funds ratio is the ratio of the amount of prepaid funds locked in the account to the total account balance; The trading frequency data includes the number of transactions per unit time and the trading interval fluctuation coefficient, which reflects the dispersion of each trading interval within a unit time. The digital currency market data includes real-time exchange rates, price volatility, and market liquidity indicators. The price volatility reflects the fluctuation range of digital currency prices within a unit of time. User social network association data includes transaction frequency of associated accounts, percentage of common trading counterparties, and abnormal social relationship markers. The percentage of common trading counterparties is the ratio of the number of common trading counterparties between the target user and associated accounts to the total number of trading counterparties of the target user.

3. The digital currency prepayment management method according to claim 1, characterized in that, The data cleaning includes: Remove missing values, outliers, and duplicate data; use interpolation to fill in missing data for key dimensions; and mark outliers that exceed a reasonable range. The normalization process is to convert data with different dimensions to a unified numerical range to eliminate the impact of differences in dimensions. The format standardization includes converting data from different sources into a preset timestamp format, monetary unit, and account identifier representation to ensure data format consistency.

4. The digital currency prepayment management method according to claim 1, characterized in that, The training process of the machine learning risk prediction model includes: Collect historical digital currency prepayment transaction data and corresponding risk event markers, divide the training set and test set, and ensure that the proportion of risk events is consistent; A hybrid model architecture combining gradient boosting trees and neural networks is adopted. The gradient boosting trees determine the feature importance weights, the neural networks output the preliminary risk probability, and the results of the two are combined to calculate the final risk probability. A specific loss function is used for model training, and cross-validation is used to avoid overfitting and adjust hyperparameters. The accuracy, recall, and F1 score of the model are calculated using the test set to verify the model's performance.

5. The digital currency prepayment management method according to claim 4, characterized in that: The feature importance weights are determined by the information gain ratio, and the weights, after normalization, sum to 1. The final risk probability is obtained by weighted summation of the risk probability output by the gradient boosting tree and the initial risk probability output by the neural network. The weighting coefficients are determined to be optimal through cross-validation combined with grid search. The hyperparameters include learning rate, tree depth, and number of hidden layer nodes in the neural network.

6. The digital currency prepayment management method according to claim 1, characterized in that, The steps for determining the risk level based on risk probability include: Clustering algorithms are used to cluster the probability distribution of historical risk events, and the initial risk level threshold is determined after obtaining the cluster centers. The initial risk levels include at least four categories: low risk, medium risk, high risk, and extremely high risk. The threshold is dynamically adjusted each preset period based on the handling effect of risk events in the previous month to ensure that the threshold is within a preset reasonable range and meets the level differentiation logic.

7. The digital currency prepayment management method according to claim 6, characterized in that, The preset risk response rules include: When the risk probability is within the low-risk threshold range, only the risk assessment result is recorded, and no proactive warning is triggered; When the risk probability is within the medium risk threshold range, a risk warning message is sent to the user to remind them to verify the transaction behavior, and a risk warning notification is pushed to the management end at the same time. When the risk probability is within the high-risk threshold range, the upper limit of the single prepayment amount is limited. The upper limit is dynamically determined based on the current available account balance and the risk probability range, and is not lower than the preset minimum prepayment amount. Warning information is also sent to the user and the management terminal simultaneously. When the risk probability exceeds the extremely high risk threshold, the digital currency prepayment function will be frozen immediately, new prepayment transactions will be prohibited, a freeze notification and risk explanation will be sent to the user, and a manual review process will be triggered on the management side. The freeze can only be lifted after the review is approved.

8. The digital currency prepayment management method according to claim 1, characterized in that, The methods for providing real-time feedback processing results include: Feedback is provided to users through multiple communication channels, including APP push, SMS, and in-app messages, with feedback delays controlled within a preset time. Provide the management with a visual interface for risk management, displaying risk events, handling measures, and user feedback in real time.

9. The digital currency prepayment management method according to claim 1, characterized in that, The risk handling log shall include at least the following contents: Risk assessment time, multi-dimensional data details, risk probability, risk level, handling measures, execution results, and user operation feedback; Logs can be filtered and retrieved by time range, account ID, risk level, etc., and the retrieval response time does not exceed the preset time. Log data is stored in a distributed manner to ensure security, and the data retention period is no less than the preset number of years.

10. A digital currency prepayment management system, characterized in that, The system is applied to the digital currency prepayment management method according to any one of claims 1-9, the system comprising: Data acquisition unit, data preprocessing unit, risk prediction unit, risk level determination unit, risk response unit, and feedback recording unit; The data acquisition unit is used to acquire multi-dimensional real-time data in the digital currency prepayment scenario; The data preprocessing unit is used to clean, normalize, and standardize the data. The risk prediction unit is used to perform feature extraction and risk probability calculation through a trained machine learning risk prediction model. The risk level determination unit is used to determine the risk level based on the risk probability; The risk response unit is used to generate early warning prompts or trigger control measures according to preset rules; The feedback recording unit is used to provide real-time feedback of processing results to the user end and the management end, and to record risk processing logs.

Images