Live broadcast e-commerce accounting management data processing system based on privacy calculation

Abstract

The invention belongs to the technical field of financial science and technology, and discloses a live broadcast e-commerce accounting management data processing system based on privacy calculation. The system comprises the following steps: collecting business operation data in real time and carrying out privacy calculation to generate privacy operation data; performing comprehensive analysis on the privacy operation data and seasonal factors, dynamically identifying an abnormal transaction mode and evaluating the credit status of a merchant; carrying out fusion analysis on the abnormal transaction mode and the credit status of the merchant, and intelligently evaluating the transaction risk; the basic financing quota of the merchant is dynamically calculated, and the basic financing quota of the merchant is dynamically adjusted based on the acquired capital risk control logic; after financing is completed, real-time monitoring and intelligent account division are carried out on the refunding fund of the merchant, and the flow direction of the refunding fund of the merchant is automatically managed; according to the method, the accuracy of the risk control model is improved, the credit risk is reduced, and the data privacy security of the merchant is ensured, so that the operation efficiency of the live broadcast e-commerce merchant and the security of the fund flow are improved.

Description

Technical Field

[0001] This invention relates to the field of financial technology, and more specifically, to a data processing system for live-streaming e-commerce accounting management based on privacy computing. Background Technology

[0002] In recent years, with the rapid development of the digital economy and the widespread adoption of mobile internet, the live-streaming e-commerce industry has experienced explosive growth. Continuous upgrades to platform technology, shifts in user consumption habits, and the powerful traffic-generating capabilities of social media have propelled live-streaming e-commerce from its initial stages to maturity, making it an indispensable force in the retail industry. However, in this rapidly evolving market environment, live-streaming e-commerce merchants face multiple operational challenges, including significant cash flow pressure, intense competition, and limited direct financing capabilities. Therefore, there is an urgent need for an intelligent live-streaming e-commerce accounting and data processing system to provide merchants with comprehensive financial service solutions, including financing and revenue sharing, helping them solve cash flow problems, improve operational efficiency, and reduce capital costs.

[0003] Patent CN109801159A discloses a blockchain-based internet e-commerce financing management method and system. The method includes: creating an internet e-commerce financing alliance chain among various banks and e-commerce platforms, wherein the alliance chain stores enterprise transaction data and credit data; creating a credit model for enterprises based on the data stored in the alliance chain; and processing financing for enterprises based on the enterprise transaction data and the credit model, thereby achieving financing management for enterprises. This invention solves the multi-party trust problem in supply chain finance scenarios and effectively reduces credit risk.

[0004] However, while the aforementioned technologies can provide financial service solutions, they primarily rely on static data modeling and on-chain data synchronization. They lack sufficient support for dynamic changes in business operations, real-time transaction risk identification, and personalized financing strategy responses. Furthermore, they lack the ability to intelligently analyze merchant behavior patterns and manage risks in a differentiated manner. This makes it difficult to detect anomalies and risk changes in business operations in a timely manner, and to respond quickly to changes in the actual operating conditions of merchants. Consequently, credit risk increases, the accuracy of risk control decreases, and the rationality of financing amounts decreases, ultimately affecting the efficiency and security of financing services.

[0005] In view of this, the present invention proposes a privacy-preserving computing-based live-streaming e-commerce accounting management data processing system to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a live-streaming e-commerce accounting management data processing system based on privacy computing, comprising:

[0007] The data acquisition module is used to collect real-time business data from merchants.

[0008] The privacy computing module is used to build a mechanism to break down data silos and to perform privacy computing on merchant operating data based on the data silo breaking down mechanism to generate privacy operating data;

[0009] The intelligent risk control module uses a pre-built AI-powered financial model to comprehensively analyze private business data and pre-extracted seasonal factors, dynamically identify abnormal transaction patterns and assess the creditworthiness of merchants; it also performs integrated analysis of abnormal transaction patterns and merchant creditworthiness to intelligently assess transaction risks.

[0010] The credit limit management module is used to integrate private business data, abnormal transaction patterns, merchant credit status and transaction risks to dynamically calculate the merchant's basic financing limit. Based on the acquired risk control logic of the funders, it provides differentiated credit limit adjustment strategies to dynamically adjust the merchant's basic financing limit and disburse loans to the merchant based on the adjusted basic financing limit.

[0011] The fund supervision module is used to build an intelligent fund supervision account system. After the financing is completed, it monitors and intelligently distributes the merchant's repayment funds in real time, and manages the flow of merchant's repayment funds automatically through smart contract technology.

[0012] Furthermore, methods for building mechanisms to break down data silos include:

[0013] A privacy-preserving computation model is built into the system. This model is an unsupervised representation learning model. The privacy-preserving computation model is sent to all e-commerce merchants. Each e-commerce merchant trains the privacy-preserving computation model locally based on its own business data, obtains the corresponding local model parameters, and uploads them to the system. The system uses a federated averaging method to aggregate all local model parameters, updates the privacy-preserving computation model based on the aggregation results, and sends the updated privacy-preserving computation model back to all e-commerce merchants for local training. This process continues until the number of times the privacy-preserving computation model is updated in the system reaches a preset threshold. At this point, the privacy-preserving computation model in the system is retrieved and marked as the global privacy-preserving computation model, thus completing the construction of the data silo breaking mechanism.

[0014] Methods for generating privacy-preserving business data include:

[0015] Input the merchant's business data into the global privacy computing model, and output the corresponding privacy-related business data.

[0016] Furthermore, methods for dynamically identifying abnormal transaction patterns and assessing merchant creditworthiness include:

[0017] Acquire merchant qualification data, assign different numerical labels to the categorical data in the merchant qualification data, and mark them as qualification labels; use a pre-trained word embedding model to convert the text data in the merchant qualification data into corresponding qualification vectors; replace each data in the merchant qualification data with the corresponding qualification label or qualification vector; use the replaced merchant qualification data, pre-extracted seasonal factors, and privacy management data as analysis data;

[0018] The analyzed data is input into a pre-built AI-powered financial and industrial model to predict corresponding pattern labels and credit scores. The AI-powered financial and industrial model includes a pattern recognition model and a credit assessment model, both of which are lightweight deep neural network models. The pattern label is a numerical label corresponding to the transaction pattern. Different transaction patterns have different pattern labels. Transaction patterns include normal transaction patterns and various abnormal transaction patterns. Abnormal transaction patterns are dynamically identified based on the pattern labels, and the credit score is used as the merchant's credit status.

[0019] Furthermore, the construction methods for AI-driven industrial finance mini-models include:

[0020] A set of 'a' sets of analytical data is collected in advance, where 'a' is an integer greater than 1. Corresponding pattern labels are assigned to each set of 'a' sets of analytical data, and the analytical data and corresponding pattern labels are converted into a set of feature vectors. Each set of feature vectors is used as input to a pattern recognition model. The pattern recognition model outputs a set of predicted pattern labels corresponding to each set of analytical data and uses the actual pattern labels corresponding to each set of analytical data as the prediction target. The actual pattern labels are the pre-set pattern labels corresponding to the analytical data. The training objective is to minimize the sum of prediction errors for all analytical data. The pattern recognition model is trained until the sum of prediction errors converges, at which point training stops, completing the construction of the pattern recognition model. The construction method for the credit assessment model is consistent with the construction method for the pattern recognition model.

[0021] Furthermore, methods for intelligently assessing transaction risks include:

[0022] If an abnormal transaction pattern is identified, multiple fuzzy sets are constructed for privacy-related business data, pattern risk level, and merchant creditworthiness. These data are then converted into membership degrees for each fuzzy set using fuzzification techniques. Fuzzy rules are defined. The fuzzified privacy-related business data, pattern risk level, and merchant creditworthiness are matched against these fuzzy rules, and fuzzy inference is performed to obtain the membership degree for each transaction risk level. The membership degrees for each transaction risk level are compared, and the transaction risk level with the highest membership degree is identified and marked as the real-time risk level. Based on the real-time risk level, the corresponding transaction risk is retrieved from a pre-defined risk matrix.

[0023] If no abnormal transaction pattern is identified, the transaction risk corresponding to the merchant's credit status is obtained from the pre-built risk mapping table. The risk mapping table consists of multiple mapping relationships, and each mapping relationship includes a merchant's credit status and the corresponding transaction risk.

[0024] Furthermore, methods for dynamically calculating a merchant's basic financing limit include:

[0025] Historical privacy data is obtained, which refers to privacy-related business data acquired at historical moments. The K-means algorithm is used to cluster the historical privacy data, obtaining c business types and their corresponding cluster centers, where c is an integer greater than 1. A type coefficient is assigned to each business type. The Euclidean distance between the privacy-related business data and each cluster center is calculated and marked as the dissimilarity. Each dissimilarity is compared, and the business type corresponding to the cluster center with the smallest dissimilarity is taken as the matching type. The type coefficient corresponding to the matching type is marked as the business coefficient.

[0026] A preset set of proportions is used, including the proportional coefficients corresponding to the risk level of the business model and the risk level of the transaction. Based on the set of proportions, the risk level of the business model and the risk level of the transaction are weighted and summed to obtain the risk coefficient. A preset set of factors is also used, including the weighted factors corresponding to the operating coefficient, the risk coefficient, and the credit coefficient. Among them, the credit coefficient represents the creditworthiness of the merchant. The operating coefficient, the risk coefficient, and the credit coefficient are multiplied by the corresponding weighted factors in the factor set to obtain the weight parameters. The weight parameters include the operating weight, the risk weight, and the credit weight. The operating weight is added to the credit weight, and then the risk weight is subtracted to obtain the merchant's basic financing amount.

[0027] Furthermore, methods for dynamically adjusting the basic financing limit for merchants include:

[0028] Based on the risk control logic of the investors, the final financing amount is generated by matching it with the creditworthiness and transaction risks of the merchants. The basic financing amount of the merchants is then dynamically adjusted based on the final financing amount.

[0029] Furthermore, methods for constructing an intelligent fund supervision account system include:

[0030] Open a fund supervision account for the merchant and bind it to the merchant's backend; set up intelligent revenue sharing strategies and smart contracts within the fund supervision account to complete the construction of the intelligent fund supervision account system.

[0031] Furthermore, methods for setting up intelligent revenue sharing strategies include:

[0032] The system retrieves the repayment progress, repayment period, and repayment amount; it then standardizes these parameters to obtain standard progress, standard period, and standard amount; a preset impact set is defined, including the impact factors corresponding to the standard progress, standard period, and standard amount; based on the impact set, a weighted sum is calculated for the standard progress, standard period, and standard amount to obtain the merchant's share; this share is subtracted from the merchant's share to obtain the repayment ratio; and the merchant's repayment funds are intelligently allocated based on the merchant's share and the repayment ratio.

[0033] Furthermore, methods for setting up smart contracts include:

[0034] The smart revenue sharing strategy is written into a smart contract using a smart contract language; the smart contract is deployed in a trusted execution environment, and a triggering mechanism for the smart contract is designed.

[0035] The technical effects and advantages of the privacy-preserving computing-based live-streaming e-commerce accounting management data processing system of this invention are as follows:

[0036] By employing privacy-preserving computation methods to construct a data silo-breaking mechanism, this system achieves privacy protection for sensitive merchant operational data and enables joint analysis and utilization across merchants. While ensuring data security and compliance, it improves the accuracy of risk control analysis and credit assessment. Integrating multi-dimensional characteristics such as privacy-preserving operational data, abnormal transaction patterns, and merchant creditworthiness, a lightweight AI-powered financial model is used to intelligently identify and assess transaction risks in real time, dynamically calculating differentiated financing amounts that match the merchant's actual operating capabilities and risk preferences. An intelligent fund supervision account system is established, using intelligent revenue sharing strategies and smart contract technology to achieve real-time monitoring and automated management of merchant repayments, effectively preventing financial risks and enhancing the transparency and credibility of fund supervision. This provides live-streaming e-commerce merchants with one-stop accounting management, financing, and fund supervision services. It not only improves the accuracy of risk control models and reduces credit risk but also protects merchants' data privacy and security, thereby improving the operational efficiency and cash flow security of live-streaming e-commerce merchants. This enables the secure utilization of live-streaming e-commerce merchants' operational data and intelligent accounting control, supporting more flexible and efficient financing solutions. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of a live-stream e-commerce accounting management data processing system based on privacy computing, according to Embodiment 1 of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Example 1

[0040] Please see Figure 1 As shown in the figure, the privacy-based computing-based live-streaming e-commerce accounting management data processing system described in this embodiment includes a data acquisition module, a privacy computing module, an intelligent risk control module, a credit limit management module, and a fund supervision module; the modules are connected by wired and / or wireless means to realize data transmission between the modules.

[0041] The data acquisition module is used to collect real-time business data from merchants.

[0042] Merchant operational data refers to the sum of various structured or unstructured information generated by e-commerce merchants during their daily operations. It objectively records the merchants' business activities, resource status, and market interactions, serving as the core basis for decision-making optimization and risk control in the live-streaming e-commerce accounting management data processing system. The live-streaming e-commerce accounting management data processing system will be referred to as the "system" below. Merchant operational data includes, but is not limited to, transaction data, financial data, and operational metrics. Transaction data includes, for example, order volume, sales revenue, and return rate; financial data includes, for example, average daily store turnover, gross profit margin, and average payment cycle; operational metrics include, for example, store rating, delivery timeliness, and positive review rate. Merchant operational data is collected through authorized sub-accounts. Specifically, e-commerce merchants add sub-accounts in their backend and authorize the system to read data. Sub-account permissions must at least cover core permissions such as product, order, and financial permissions to ensure the completeness and accuracy of the collected data.

[0043] The privacy computing module is used to build a mechanism to break down data silos and perform privacy computing on merchant operating data based on the data silo breaking down mechanism to generate privacy-preserving operating data.

[0044] Methods for building a data silo-breaking mechanism include:

[0045] A privacy-preserving computation model is constructed in the system. This model is an unsupervised representation learning model, such as an autoencoder, variational autoencoder, or feature compressor. Unsupervised representation learning models are existing technologies, and the specific construction process will not be elaborated here. The privacy-preserving computation model is sent to all e-commerce merchants. Each e-commerce merchant trains the privacy-preserving computation model locally based on its corresponding business data, obtains the corresponding local model parameters, and uploads them to the system. The system uses a federated averaging method to aggregate all local model parameters, updates the privacy-preserving computation model based on the aggregation results, and sends the updated privacy-preserving computation model back to all e-commerce merchants for local training. This process continues until the number of times the privacy-preserving computation model is updated in the system reaches a preset threshold. At this point, the privacy-preserving computation model in the system is acquired and marked as the global privacy-preserving computation model, thus completing the construction of the data silo breaking mechanism. It should be noted that local training of the unsupervised representation learning model and the federated averaging method are existing technologies, and the specific process will not be elaborated here. The threshold number of updates is preset by those skilled in the art based on the actual situation.

[0046] It should be understood that the reason for building a data silo-breaking mechanism is that the operational data of e-commerce merchants are scattered across different platforms, forming data silos, which limits the sharing and comprehensive utilization of data and affects the effectiveness of key businesses such as risk control and credit assessment. Furthermore, due to data privacy and compliance requirements, it is not possible to directly and centrally store merchants' operational data. Therefore, there is an urgent need for a mechanism that can both enable collaborative utilization of data from multiple merchants and ensure data privacy and security, namely, a data silo-breaking mechanism.

[0047] The reason for using unsupervised representation learning models to build a data silo-breaking mechanism is that unsupervised representation learning models can extract privacy features and effectively represent merchants' operational data; by distributing unsupervised representation learning models to each merchant's local machine for training, the leakage of raw data is avoided, ensuring privacy and security; by aggregating local model parameters through federated averaging, collaborative modeling of scattered data is achieved; and multiple rounds of iterative training continuously optimize the unsupervised representation learning model, thereby breaking down data silos while protecting privacy and enabling joint utilization of cross-merchant data.

[0048] Methods for generating privacy-preserving business data include:

[0049] Merchant operating data is input into the global privacy computing model, and the corresponding privacy operating data is output. Privacy operating data refers to the data representation generated by the global privacy computing model after privacy transformation or encryption mapping of merchant operating data. It has the characteristics of irreversibility, desensitization and secure computing, and can be used in business scenarios such as risk control analysis and credit assessment, effectively avoiding the exposure of the merchant's original sensitivity.

[0050] The intelligent risk control module uses a pre-built AI-powered financial model to comprehensively analyze private business data and pre-extracted seasonal factors, dynamically identify abnormal transaction patterns and assess the creditworthiness of merchants; it also performs integrated analysis of abnormal transaction patterns and merchant creditworthiness to intelligently assess transaction risks.

[0051] Methods for dynamically identifying abnormal transaction patterns and assessing merchant creditworthiness include:

[0052] Merchant qualification data is obtained, including but not limited to business categories, business operation status, corporate credit, and legal representative credit, through a comprehensive approach including business registration information, third-party corporate information query platforms, official credit reporting systems (such as the People's Bank of China Credit Reference Center and the People's Bank of China Enterprise Credit Reference Center), and third-party credit reporting agency queries. Different numerical labels are assigned to categorical data (such as business categories and business operation status) within the merchant qualification data, and these are marked as qualification labels. A pre-trained word embedding model (such as the Word2Vec model or the GloVe model) is used to convert textual data (such as corporate credit and legal representative credit) within the merchant qualification data into corresponding qualification vectors. The word embedding model is an existing technology, and the specific training process will not be elaborated upon here.

[0053] Each data point in the merchant qualification data is replaced with its corresponding qualification tag or qualification vector. The replaced merchant qualification data, pre-extracted seasonal factors, and privacy-related business data are used as analysis data. Seasonal factors refer to the regular changes in merchant business data within a specific time period (such as month, quarter, holidays, etc.), including but not limited to holiday effects (such as a surge in order volume and sales during e-commerce promotions like Double 11 and 618; festivals like Spring Festival and Mid-Autumn Festival may also affect consumer purchasing behavior) and quarterly changes (such as significant differences in demand for certain categories in different quarters, with distinct peak and off-seasons). Seasonal factors are obtained by those skilled in the art through time series analysis and feature extraction of merchant business data acquired at historical moments. It should be understood that the pre-extraction of seasonal factors as auxiliary features helps the AI ​​financial model distinguish between normal cyclical fluctuations and abnormal transaction patterns, avoiding misjudging normal fluctuations as abnormal transaction patterns, thereby improving the accuracy and stability of subsequent transaction risk assessment.

[0054] The analyzed data is input into a pre-built AI-powered financial and industrial model to predict corresponding pattern labels and credit scores. This model includes a pattern recognition model and a credit assessment model, both of which are lightweight deep neural network models. The pattern recognition model predicts pattern labels, while the credit assessment model predicts credit scores. Pattern labels are numerical labels corresponding to different transaction patterns. Transaction patterns include normal transaction patterns and various abnormal transaction patterns, such as sudden changes in transaction amount, high-frequency small-amount transactions, and abnormal order return rates. Abnormal transaction patterns are dynamically identified based on the pattern labels, and the credit score is used to determine the merchant's creditworthiness.

[0055] The methods for constructing AI-powered industrial and financial mini-models include:

[0056] A set of analysis data (a) is collected in advance, where a is an integer greater than 1. Corresponding pattern labels are set for each set of analysis data (a). The analysis data and corresponding pattern labels are converted into a set of feature vectors. The pattern labels corresponding to the analysis data are collected by a person skilled in the art during the historical identification of abnormal transaction patterns. The analysis data (a) is collected, and each set of analysis data is analyzed according to the actual situation to identify the corresponding transaction patterns. Corresponding pattern labels are set for the analysis data (a) in sequence.

[0057] Each set of feature vectors is used as input to the pattern recognition model. The pattern recognition model outputs a set of predicted pattern labels corresponding to each set of analyzed data, and uses the actual pattern labels corresponding to each set of analyzed data as the prediction target. The actual pattern labels are the pre-set pattern labels corresponding to the analyzed data. The training objective is to minimize the sum of prediction errors for all analyzed data. The prediction error is calculated using the formula η. b =(θ b -ε b ) 2 , where η b Let b be the prediction error, b be the group number of the feature vector corresponding to the analyzed data, and θ be the prediction error. b ε represents the predicted pattern label corresponding to the b-th group of analyzed data. b The actual pattern label corresponds to the b-th group of analyzed data; the pattern recognition model is trained until the sum of prediction errors converges, at which point training stops, thus completing the construction of the pattern recognition model; the construction method of the credit assessment model is the same as that of the pattern recognition model.

[0058] Methods for intelligently assessing transaction risk include:

[0059] If abnormal transaction patterns are identified, multiple fuzzy sets are constructed for privacy-related business data, pattern risk levels, and merchant creditworthiness. For example, the fuzzy sets corresponding to pattern risk levels include high risk, medium risk, and low risk. The privacy-related business data, pattern risk levels, and merchant creditworthiness are then converted into membership degrees for each corresponding fuzzy set using fuzzification techniques. Fuzzification is the process of converting precise numerical values ​​into membership degrees corresponding to fuzzy sets; fuzzification techniques include triangular membership functions and trapezoidal membership functions. Fuzzy rules are defined based on expert knowledge or relevant literature. The fuzzified privacy-related business data, pattern risk levels, and merchant creditworthiness are then compared with... Fuzzy rule matching is performed, and fuzzy inference methods (such as the Mamdani fuzzy inference model, the Sugeno fuzzy inference model, etc.) are used to obtain fuzzy inference results. The fuzzy inference results are the membership degrees of each transaction risk level, including low risk, medium risk, and high risk. The membership degrees of each transaction risk level are compared, and the transaction risk level with the highest membership degree is obtained and marked as the real-time risk level. Based on the real-time risk level, the corresponding transaction risk is obtained from a preset risk matrix. The risk matrix includes the transaction risks corresponding to different transaction risk levels, and the transaction risks are specific numerical values. The risk matrix is ​​preset by those skilled in the art according to the actual situation.

[0060] If no abnormal transaction pattern is identified, the transaction risk corresponding to the merchant's creditworthiness is obtained from the pre-built risk mapping table. The risk mapping table consists of multiple sets of mapping relationships, each of which includes a merchant's creditworthiness and the corresponding transaction risk. The risk mapping table is pre-built by those skilled in the art based on the actual situation.

[0061] It should be noted that the anomaly intensity, pattern risk level, and merchant creditworthiness characterize transaction risk from three dimensions: behavioral deviation, risk type, and entity credit. The anomaly intensity reflects the degree to which the transaction behavior deviates from the merchant's normal business model and is a direct signal of risk occurrence. The pattern risk level represents the known abnormal transaction pattern and reflects its inherent risk level. The merchant creditworthiness reflects the overall credit reliability of the merchant and affects the degree to which abnormal transaction patterns are tolerated or guarded against. The anomaly intensity, pattern risk level, and merchant creditworthiness complement each other from different perspectives, enabling a more comprehensive and accurate assessment of whether a transaction carries risk.

[0062] The credit limit management module integrates private business data, abnormal transaction patterns, merchant credit status, and transaction risks to dynamically calculate the merchant's basic financing limit. Based on the acquired risk control logic of the funding party, it provides differentiated credit limit adjustment strategies to dynamically adjust the merchant's basic financing limit and disburse loans to the merchant based on the adjusted basic financing limit.

[0063] Methods for dynamically calculating a merchant's basic financing limit include:

[0064] Historical privacy data is acquired, which refers to privacy-related business data obtained at historical moments. The K-means algorithm is used to cluster this historical privacy data, resulting in c business types and their corresponding cluster centers, where c is an integer greater than 1. Each cluster center is a core representative point of its business type, essentially the mean vector of all data within that type, representing the typical characteristic vector of that type and reflecting its common state. Business types include, for example, startups, high-potential businesses, and mature, stable businesses. Those skilled in the art can set corresponding type coefficients for different business types based on the specific circumstances. The Euclidean distance between the privacy-related business data and each cluster center is calculated and marked as the degree of difference. Each degree of difference is compared, and the business type corresponding to the cluster center with the smallest degree of difference is taken as the matching type. The type coefficient corresponding to the matching type is marked as the business coefficient. The calculation method for the Euclidean distance and the K-means algorithm are existing technologies, and the specific process will not be elaborated upon here.

[0065] A preset set of proportions is used, which includes proportional coefficients corresponding to the risk level of the business model and the risk level of the transaction. This set of proportions is pre-set by those skilled in the art based on actual conditions. Based on this set of proportions, the risk level of the business model and the risk level of the transaction are weighted and summed to obtain a risk coefficient. A preset set of factors is also used, which includes weighted factors corresponding to the operating coefficient, risk coefficient, and credit coefficient. This set of factors is pre-constructed by those skilled in the art based on actual conditions. The credit coefficient represents the creditworthiness of the merchant. The operating coefficient, risk coefficient, and credit coefficient are multiplied by their corresponding weighted factors in the factor set to obtain weight parameters. These weight parameters include operating weight, risk weight, and credit weight. The operating weight is added to the credit weight, and then the risk weight is subtracted to obtain the merchant's basic financing amount.

[0066] It should be understood that private operational data, abnormal transaction patterns, merchant creditworthiness, and transaction risks collectively reflect a merchant's true operating status and financing security from multiple dimensions, including operational capabilities, transaction stability, credit level, and risk control. Private operational data reflects a merchant's actual operational level and is the foundation of financing capabilities; abnormal transaction patterns reveal potential operational anomalies, affecting the judgment of the compliance of fund usage; a merchant's creditworthiness reflects their willingness and ability to repay, and is the core of credit assessment; transaction risks comprehensively assess the security and stability of the overall cash flow, directly affecting the expected lending risk; a comprehensive analysis of private operational data, abnormal transaction patterns, merchant creditworthiness, and transaction risks helps to accurately and dynamically assess and determine a merchant's basic financing amount.

[0067] Methods for dynamically adjusting a merchant's basic financing limit include:

[0068] Based on the acquired investor risk control logic, the system matches the merchant's creditworthiness and transaction risk to generate a final financing limit. The merchant's basic financing limit is then dynamically adjusted based on this final limit. The investor risk control logic refers to a logical model that, during the credit limit management process, personalizes and dynamically adjusts the merchant's basic financing limit according to the risk control strategies and credit rules set by the investor (such as banks or financial institutions). This model primarily relies on the merchant's creditworthiness and transaction risk assessment results, combined with factors such as the investor's risk tolerance, credit policies, and industry preferences, to form differentiated credit limit adjustment strategies, achieving flexible credit granting and precise risk control. For example, the investor risk control logic could be set as follows: when a merchant's creditworthiness is below 80, their basic financing limit is reduced by 30%; when a merchant's transaction risk is less than or equal to 0.3, their basic financing limit remains unchanged. The investor risk control logic is directly provided by the investor.

[0069] It's important to note that the reason for calculating the basic financing limit before dynamic adjustment is that the basic financing limit represents the initial creditworthiness based on objective factors such as the merchant's actual operating data, abnormal transaction patterns, creditworthiness, and transaction risks. This reflects the theoretical value of financing for the merchant without external intervention. Based on this, and considering the different risk control logics of different investors, the basic limit is dynamically adjusted in a differentiated manner. This fully considers the investor's risk tolerance, industry preferences, and specific risk control requirements, enabling personalized and refined management of the financing limit. This not only ensures the transparency and scientific nature of the limit calculation but also enhances the system's flexibility and scalability, facilitating support for complex business scenarios with multiple investors and strategies. Simultaneously, it ensures that the financing limit aligns with both the merchant's actual operating capabilities and the investor's risk management needs, achieving flexible, dynamic, and precise financing limit management.

[0070] The fund supervision module is used to build an intelligent fund supervision account system. After the financing is completed, it monitors and intelligently distributes the merchant's repayment funds in real time, and manages the flow of merchant's repayment funds automatically through smart contract technology.

[0071] Methods for constructing an intelligent fund supervision account system include:

[0072] A fund supervision account is opened for merchants and linked to their back-end system. This account serves as the withdrawal and repayment account, used to uniformly collect and manage merchants' repayment funds. A fund supervision account is a dedicated account established by the platform, financial institution, or third party during specific transactions, business operations, or financing processes to ensure the safety and compliance of fund flows. It is used for full-process monitoring, restriction of use, and automatic allocation of funds. Merchant repayment funds refer to the actual payments or revenue received by merchants from customers or the platform after selling goods or services. Within the fund supervision account, intelligent revenue sharing strategies and smart contracts are set up to complete the construction of the intelligent fund supervision account system. The intelligent revenue sharing rules are used to automatically allocate each merchant repayment (e.g., repayment of financing, platform deductions, merchant withdrawable balance, etc.), and the smart contract automatically triggers the execution of the revenue sharing results, ensuring that the entire fund allocation process is transparent, efficient, traceable, and tamper-proof, thereby achieving a closed-loop construction and stable operation of the fund supervision account system.

[0073] Methods for setting up intelligent revenue sharing strategies include:

[0074] Obtain repayment progress, repayment period, and repayment amount; repayment progress is the percentage of the repaid financing amount in the final financing amount, obtained from the repayment records in the fund supervision account; the repayment period is the remaining time between the current time and the repayment deadline stipulated in the financing contract; the repayment amount is the total amount of funds recovered by the merchant, obtained from the fund supervision account; standardize the repayment progress, repayment period, and repayment amount sequentially (e.g., mean-variance standardization, maximum absolute value standardization, etc.) to obtain standard progress, standard period, and standard amount; preset the influence set, the influence set... The mechanism includes influencing factors corresponding to standard progress, standard deadline, and standard amount. The influencing set is pre-set by those skilled in the art based on actual circumstances. Based on the influencing set, the standard progress, standard deadline, and standard amount are weighted and summed to obtain the merchant's share. The merchant's share is then subtracted from the sum to obtain the repayment ratio. The repayment ratio is the proportion of the merchant's recovered funds used to repay financing (i.e., return funds to the investor), while the merchant's share is the proportion of the merchant's recovered funds remaining in the merchant's own account (or balance) for operation and withdrawal. The merchant's recovered funds are intelligently allocated based on the merchant's share and the repayment ratio.

[0075] Methods for setting up smart contracts include:

[0076] Using smart contract languages ​​such as Solidity, Rust, and Move, the smart revenue sharing strategy is written as a code-based smart contract to ensure its automatic execution. The smart contracts are deployed in a trusted execution environment, such as a blockchain network (Ethereum, consortium blockchains, etc.) or a trusted execution module within the system (i.e., a contract execution environment with trusted computing and automatic execution capabilities implemented within the platform's own architecture, relying on a public blockchain; typically used to implement similar smart contract functions while offering higher execution efficiency and business controllability), to ensure the immutability of the smart revenue sharing strategy's execution process. A trigger mechanism for the smart contracts is designed to ensure the smart revenue sharing strategy executes automatically at appropriate times; trigger mechanisms include automatic triggering when merchant payments are received or triggering after a certain repayment amount is reached.

[0077] It should be noted that the purpose of building an intelligent fund supervision account system is to uniformly collect the funds returned by merchants after they have completed financing and transfer them to the supervision account. By combining intelligent revenue sharing strategies and smart contracts, the system enables full-process monitoring and automated allocation of the returned funds, thereby ensuring the safety of the investor's funds and preventing the risk of merchant default. At the same time, it ensures that the flow of funds is transparent, traceable, and tamper-proof, meeting compliance and regulatory requirements. In addition, it can also realize personalized revenue sharing strategies based on dynamic factors such as the merchant's repayment progress and the amount of returned funds, improving the efficiency and flexibility of fund management and building a high-trust, high-risk control cooperation mechanism between the platform and the investor.

[0078] This embodiment employs privacy-preserving computation methods to construct a data silo-breaking mechanism, achieving privacy protection for sensitive merchant operational data and cross-merchant joint analysis and utilization. While ensuring data security and compliance, it improves the accuracy of risk control analysis and credit assessment. Integrating multi-dimensional features such as privacy-preserving operational data, abnormal transaction patterns, and merchant creditworthiness, it uses a lightweight AI-powered financial model to intelligently identify and assess transaction risks in real time, dynamically calculating differentiated financing amounts that match the merchant's actual operating capabilities and risk preferences. An intelligent fund supervision account system is established, using intelligent revenue sharing strategies and smart contract technology to achieve real-time monitoring and automated management of merchant repayments, effectively preventing financial risks and enhancing the transparency and credibility of fund supervision. It provides live-streaming e-commerce merchants with one-stop accounting management, financing, and fund supervision services, not only improving the accuracy of risk control models and reducing credit risk but also protecting merchants' data privacy and security. This enhances the operational efficiency and cash flow security of live-streaming e-commerce merchants, enabling the secure utilization of their operational data and intelligent accounting control, supporting more flexible and efficient financing solutions.

[0079] Example 2

[0080] This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform a privacy-based computing-based live-streaming e-commerce accounting data processing system as described above.

[0081] The methods or systems according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store a privacy-based computing-based live-streaming e-commerce accounting management data processing system provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components of the electronic device shown in this application may be omitted according to actual needs.

[0082] Example 3

[0083] One embodiment of this application discloses a computer-readable storage medium. The computer-readable storage medium stores computer-readable instructions. When the computer-readable instructions are executed by a processor, they can perform a privacy-based computing-based live-streaming e-commerce accounting data processing system according to an embodiment of this application, as described with reference to the above figures. The storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0084] Furthermore, according to embodiments of this application, the processes described in the above-referenced flowcharts can be implemented as computer software programs. For example, this application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be executed by a processor to perform instructions corresponding to the method steps provided in this application, such as a privacy-based computing-based live-streaming e-commerce accounting management data processing system. When this computer program is executed by a central processing unit (CPU), it performs the functions defined in the method of this application.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0086] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0087] In the description of this invention, it should be understood that the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0088] In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0089] In the description of this invention, "several" means one or more, and "a large number" means two or more.

[0090] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0091] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0092] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A data processing system for live-streaming e-commerce accounting management based on privacy computing, characterized in that, include: The data acquisition module is used to collect real-time business data from merchants. The privacy computing module is used to build a mechanism to break down data silos and to perform privacy computing on merchant operating data based on the data silo breaking down mechanism to generate privacy operating data; The intelligent risk control module uses a pre-built AI-powered financial model to comprehensively analyze private business data and pre-extracted seasonal factors, dynamically identify abnormal transaction patterns and assess the creditworthiness of merchants; it also performs integrated analysis of abnormal transaction patterns and merchant creditworthiness to intelligently assess transaction risks. The credit limit management module is used to integrate private business data, abnormal transaction patterns, merchant credit status and transaction risks to dynamically calculate the merchant's basic financing limit. Based on the acquired risk control logic of the funders, it provides differentiated credit limit adjustment strategies to dynamically adjust the merchant's basic financing limit and disburse loans to the merchant based on the adjusted basic financing limit. The fund supervision module is used to build an intelligent fund supervision account system. After the financing is completed, it monitors and intelligently distributes the merchant's repayment funds in real time, and manages the flow of merchant's repayment funds automatically through smart contract technology.

2. The data processing system for live-streaming e-commerce accounting management based on privacy computing as described in claim 1, characterized in that, Methods for building a data silo-breaking mechanism include: A privacy-preserving computation model is built into the system. This model is an unsupervised representation learning model. The privacy-preserving computation model is sent to all e-commerce merchants. Each e-commerce merchant trains the privacy-preserving computation model locally based on its own business data, obtains the corresponding local model parameters, and uploads them to the system. The system uses a federated averaging method to aggregate all local model parameters, updates the privacy-preserving computation model based on the aggregation results, and sends the updated privacy-preserving computation model back to all e-commerce merchants for local training. This process continues until the number of times the privacy-preserving computation model is updated in the system reaches a preset threshold. At this point, the privacy-preserving computation model in the system is retrieved and marked as the global privacy-preserving computation model, thus completing the construction of the data silo breaking mechanism. Methods for generating privacy-preserving business data include: Input the merchant's business data into the global privacy computing model, and output the corresponding privacy-related business data.

3. The data processing system for live-streaming e-commerce accounting management based on privacy computing according to claim 2, characterized in that, Methods for dynamically identifying abnormal transaction patterns and assessing merchant creditworthiness include: Acquire merchant qualification data, assign different numerical labels to the categorical data in the merchant qualification data, and mark them as qualification labels; use a pre-trained word embedding model to convert the text data in the merchant qualification data into corresponding qualification vectors; replace each data in the merchant qualification data with the corresponding qualification label or qualification vector; use the replaced merchant qualification data, pre-extracted seasonal factors, and privacy management data as analysis data; The analyzed data is input into a pre-built AI-powered financial and industrial model to predict corresponding pattern labels and credit scores. The AI-powered financial and industrial model includes a pattern recognition model and a credit assessment model, both of which are lightweight deep neural network models. The pattern label is a numerical label corresponding to the transaction pattern. Different transaction patterns have different pattern labels. Transaction patterns include normal transaction patterns and various abnormal transaction patterns. Abnormal transaction patterns are dynamically identified based on the pattern labels, and the credit score is used as the merchant's credit status.

4. The data processing system for live-streaming e-commerce accounting management based on privacy computing according to claim 3, characterized in that, The methods for constructing AI-powered industrial and financial mini-models include: A set of 'a' sets of analytical data is collected in advance, where 'a' is an integer greater than 1. Corresponding pattern labels are assigned to each set of 'a' sets of analytical data, and the analytical data and corresponding pattern labels are converted into a set of feature vectors. Each set of feature vectors is used as input to a pattern recognition model. The pattern recognition model outputs a set of predicted pattern labels corresponding to each set of analytical data and uses the actual pattern labels corresponding to each set of analytical data as the prediction target. The actual pattern labels are the pre-set pattern labels corresponding to the analytical data. The training objective is to minimize the sum of prediction errors for all analytical data. The pattern recognition model is trained until the sum of prediction errors converges, at which point training stops, completing the construction of the pattern recognition model. The construction method for the credit assessment model is consistent with the construction method for the pattern recognition model.

5. The data processing system for live-streaming e-commerce accounting management based on privacy computing according to claim 4, characterized in that, Methods for intelligently assessing transaction risk include: If an abnormal transaction pattern is identified, multiple fuzzy sets are constructed for privacy-related business data, pattern risk level, and merchant creditworthiness. These data are then converted into membership degrees for each fuzzy set using fuzzification techniques. Fuzzy rules are defined. The fuzzified privacy-related business data, pattern risk level, and merchant creditworthiness are matched against these fuzzy rules, and fuzzy inference is performed to obtain the membership degree for each transaction risk level. The membership degrees for each transaction risk level are compared, and the transaction risk level with the highest membership degree is identified and marked as the real-time risk level. Based on the real-time risk level, the corresponding transaction risk is retrieved from a pre-defined risk matrix. If no abnormal transaction pattern is identified, the transaction risk corresponding to the merchant's credit status is obtained from the pre-built risk mapping table. The risk mapping table consists of multiple mapping relationships, and each mapping relationship includes a merchant's credit status and the corresponding transaction risk.

6. The data processing system for live-streaming e-commerce accounting management based on privacy computing according to claim 5, characterized in that, Methods for dynamically calculating a merchant's basic financing limit include: Historical privacy data is obtained, which refers to privacy-related business data acquired at historical moments. The K-means algorithm is used to cluster the historical privacy data, obtaining c business types and their corresponding cluster centers, where c is an integer greater than 1. A type coefficient is assigned to each business type. The Euclidean distance between the privacy-related business data and each cluster center is calculated and marked as the dissimilarity. Each dissimilarity is compared, and the business type corresponding to the cluster center with the smallest dissimilarity is taken as the matching type. The type coefficient corresponding to the matching type is marked as the business coefficient. A preset set of proportions is used, including the proportional coefficients corresponding to the risk level of the business model and the risk level of the transaction. Based on the set of proportions, the risk level of the business model and the risk level of the transaction are weighted and summed to obtain the risk coefficient. A preset set of factors is also used, including the weighted factors corresponding to the operating coefficient, the risk coefficient, and the credit coefficient. Among them, the credit coefficient represents the creditworthiness of the merchant. The operating coefficient, the risk coefficient, and the credit coefficient are multiplied by the corresponding weighted factors in the factor set to obtain the weight parameters. The weight parameters include the operating weight, the risk weight, and the credit weight. The operating weight is added to the credit weight, and then the risk weight is subtracted to obtain the merchant's basic financing amount.

7. The live-streaming e-commerce accounting management data processing system based on privacy computing according to claim 6, characterized in that, Methods for dynamically adjusting a merchant's basic financing limit include: Based on the risk control logic of the investors, the final financing amount is generated by matching it with the creditworthiness and transaction risks of the merchants. The basic financing amount of the merchants is then dynamically adjusted based on the final financing amount.

8. The live-stream e-commerce accounting management data processing system based on privacy computing according to claim 7, characterized in that, Methods for constructing an intelligent fund supervision account system include: Open a fund supervision account for the merchant and bind it to the merchant's backend; set up intelligent revenue sharing strategies and smart contracts within the fund supervision account to complete the construction of the intelligent fund supervision account system.

9. A data processing system for live-streaming e-commerce accounting management based on privacy computing as described in claim 8, characterized in that, Methods for setting up intelligent revenue sharing strategies include: The system retrieves the repayment progress, repayment period, and repayment amount; it then standardizes these parameters to obtain standard progress, standard period, and standard amount; a preset impact set is defined, including the impact factors corresponding to the standard progress, standard period, and standard amount; based on the impact set, a weighted sum is calculated for the standard progress, standard period, and standard amount to obtain the merchant's share; this share is subtracted from the merchant's share to obtain the repayment ratio; and the merchant's repayment funds are intelligently allocated based on the merchant's share and the repayment ratio.

10. A data processing system for live-streaming e-commerce accounting management based on privacy computing as described in claim 9, characterized in that, Methods for setting up smart contracts include: The smart revenue sharing strategy is written into a smart contract using a smart contract language; the smart contract is deployed in a trusted execution environment, and a triggering mechanism for the smart contract is designed.

Images