Cross-border bulk commodity e-commerce transaction risk assessment management system

Abstract

The invention discloses a cross-border bulk commodity e-commerce transaction risk assessment management system, which relates to the technical field of e-commerce and comprises a transaction risk management center. The transaction risk management center is in communication connection with a data acquisition and integration module, a knowledge graph module, a risk trend analysis module, a risk early warning module, a countermeasure recommendation module and a risk assessment report generation module, and all the modules are in electric signal connection. According to the method, the risk identification model is constructed, the risk knowledge and the association relationship in the knowledge graph of the transaction behavior are combined, the risk identification is carried out on the early stage, the middle stage and the later stage of the transaction behavior, and meanwhile, the corresponding risk early warning threshold values are set for different periods of the transaction behavior to trigger the early warning signal. An enterprise can take measures before the risk occurs or at the initial stage, so that the loss caused by the risk is effectively avoided or reduced, and the coping capacity of the enterprise to the risk is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce, and in particular to a risk assessment and management system for cross-border bulk commodity e-commerce transactions. Background Art

[0002] With the advancement of global economic integration, the scale of cross-border bulk commodity transactions continues to expand. Commodities are characterized by high value, large trading volume, and frequent price fluctuations. Since cross-border bulk commodity trade involves multiple links, including procurement, transportation, warehousing, settlement, etc., and the transaction process is relatively long, in e-commerce transactions, it is highly dependent on the Internet and information technology. Network security issues have become one of the important challenges facing cross-border bulk commodity e-commerce transactions.

[0003] For example, China Patent Publication No. CN109670736A discloses a risk management method for e-commerce transactions, which implements unified management of e-commerce transactions by setting up a risk management system for e-commerce transactions, adopts an e-commerce app based on the risk management system for e-commerce transactions, establishes a secure data channel between transaction users and system servers, and realizes unified management of e-commerce transactions.

[0004] In the prior art, a risk management system based on e-commerce transaction behavior is used to uniformly manage and supervise e-commerce transaction behavior, which solves the problem of chaotic e-commerce transactions and uncontrollable risk management. However, since transaction risks exist not only after the transaction, but also before and during the transaction, passive assessments are performed after the transaction occurs, and there is a lack of an effective early warning mechanism, which affects the security of transaction behavior. Therefore, how to combine the risk knowledge and association relationships of the knowledge graph to evaluate the risk trends of the early, middle and late stages of the transaction behavior, identify potential risk signs, and issue early warnings so that countermeasures can be taken is the problem to be solved by the present invention. To this end, a cross-border bulk commodity e-commerce transaction risk assessment and management system is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide a cross-border bulk commodity e-commerce transaction risk assessment and management system to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A cross-border bulk commodity e-commerce transaction risk assessment and management system includes a transaction risk management center, wherein the transaction risk management center is communicatively connected to a data collection and integration module, a knowledge graph module, a risk trend analysis module, a risk warning module, a response measure recommendation module, and a risk assessment report generation module, wherein the modules are electrically connected;

[0008] The transaction risk management center is responsible for coordinating the communication and data interaction between various modules, ensuring close coordination between all links and improving the coordination and efficiency of the entire system operation;

[0009] The data collection and integration module is used to collect data related to cross-border commodity transactions from multiple data sources, and pre-process and integrate the collected data to obtain a basic transaction data set;

[0010] The knowledge graph module is used to construct a knowledge graph of transaction behavior, integrate various types of information involved in the transaction, including suppliers, buyers, logistics providers, market conditions, and historical transaction records, screen out risk features related to transaction risk assessment, and form a risk feature sequence list;

[0011] The risk trend analysis module is used to construct a risk identification model based on the risk features in the risk feature sequence table, identify risks at the early, middle, and late stages of a transaction, and analyze the risk trend of the transaction period based on the deviation of each risk feature from its own preset benchmark value;

[0012] The risk warning module is used to conduct risk assessment of trading behaviors based on the results of risk identification and risk trend analysis, set corresponding risk warning thresholds for different periods of trading behaviors, trigger warning signals, and issue risk warning information;

[0013] The response measure recommendation module is used to combine a pre-established response strategy library containing various risk response strategies, based on the risk warning information issued by the risk warning module, and in combination with the transaction-related party information and industry knowledge in the knowledge graph, to intelligently recommend appropriate response measures from the response strategy library;

[0014] The risk assessment report generation module is used to automatically generate a cross-border bulk commodity e-commerce transaction risk assessment report according to a set period.

[0015] A further improvement of the technical solution of the present invention is that the data acquisition and integration module specifically includes:

[0016] Integrate multiple interfaces to collect data related to cross-border bulk commodity transactions from multiple data sources, including cross-border e-commerce platforms, third-party payment platforms, logistics companies, and external market data, including but not limited to order information, payment information, logistics information, supplier and buyer credit data, market price fluctuation data, etc. Specifically, it captures order information from cross-border e-commerce platforms, obtains payment information from third-party payment platforms, connects with logistics company systems to collect logistics information, and simultaneously obtains supplier and buyer credit data and market price fluctuation data from external market data sources and internal credit assessment systems to ensure that data comprehensively covers all aspects of cross-border bulk commodity transactions;

[0017] Pre-processing of collected data related to cross-border commodity transactions, including data cleaning, outlier processing, and standardization;

[0018] The different types of pre-processed data are associated and integrated according to the fields of the transaction order number, and the order information, payment information, logistics information, credit data and market price fluctuation data are integrated to form a complete and coherent transaction record. Through data mapping and conversion, the basic transaction data set is constructed.

[0019] A further improvement of the technical solution of the present invention is that the knowledge graph module specifically includes:

[0020] The basic transaction data set is obtained from the data collection and integration module. Natural language processing technology is used to parse the text data involved. In combination with the rule engine, based on the preset rule template, the entities in the basic transaction data set, including supplier names, buyer information, logistics order numbers, etc., are identified and extracted. The association relationship between the entities is established, and the scattered data points are integrated into a structured information network. At the same time, the attribute information in the data is extracted to build a preliminary knowledge graph structure.

[0021] Based on the actual needs of cross-border commodity trading risk assessment, risk characteristics related to transaction risk assessment are screened from the integrated information, and trading behavior is divided into three periods: early stage, mid-term, and late stage. The key risk characteristics of each stage of trading behavior are clearly defined. Among them, early stage risk characteristics of trading behavior include counterparty credit score, historical transaction default rate, and supply and demand change rate. Mid-term risk characteristics of trading behavior include price volatility, cargo transportation delay rate, and number of days of delay in funds arrival. Late stage risk characteristics of trading behavior include the number of transaction disputes, product delivery quality qualification rate, and accounts receivable collection period.

[0022] The screened risk features are classified and organized according to the early, middle and late stages of the transaction behavior to form a risk feature sequence table. The risk feature sequence table clearly lists the risk features of each stage, assigns corresponding identification and attribute descriptions to each risk feature, and maps the risk feature sequence table to the knowledge graph to form a subgraph structure containing risk feature nodes and association relationships.

[0023] A further improvement of the technical solution of the present invention is that: the risk trend analysis module includes a risk identification model construction unit and a risk identification unit;

[0024] The risk identification model construction unit is used to utilize the historical transaction data stored in the knowledge graph and, in combination with the risk features related to transaction risk assessment in the risk feature sequence table, to construct a risk identification model based on a neural network model;

[0025] The risk identification unit is used to use the constructed risk identification model to identify risks in different periods of trading behavior, analyze the deviation of each risk characteristic from its respective benchmark value, calculate the early risk trend index, the mid-term risk trend index and the late risk trend index, and clarify the risk trend of each period of trading behavior.

[0026] A further improvement of the technical solution of the present invention is that the risk identification model building unit specifically includes:

[0027] Extract historical transaction data related to transaction risk assessment from the knowledge graph, and screen out risk feature data related to transaction risk assessment based on the risk feature types covered by the risk feature sequence table. Normalize the features, scaling numerical features to the range of 0 to 1 to eliminate dimensional differences. At the same time, based on user needs and historical transaction data, preset baseline values for each risk feature, mark risk features that deviate from the baseline values, and integrate them into a comprehensive feature dataset.

[0028] Based on the characteristics and requirements of cross-border commodity trading risk assessment, a neural network architecture was selected to build a multi-layer perceptron (MLP) network model, which includes an input layer, multiple hidden layers, and an output layer.

[0029] The prepared feature data set is divided into a training set and a validation set. The constructed neural network model is trained using the training set. The model training process is adjusted by setting the hyperparameters of the learning rate and the number of iterations. During the training process, the loss function is calculated using the backpropagation algorithm, and the network weights are adjusted to minimize the loss function value so that the model learns the mapping relationship between risk characteristics and transaction risks. During the training process, the model is verified using the validation set to evaluate the performance of the model on unseen data and prevent the model from overfitting. By adjusting the hyperparameters and optimizing the network structure, the model can achieve better performance on the validation set, enabling the model to accurately identify risks in cross-border bulk commodity transactions, until the model achieves satisfactory performance and obtains the final risk identification model.

[0030] A further improvement of the technical solution of the present invention is that the risk identification unit specifically includes:

[0031] The pre-processed risk feature data is input into the established risk identification model and compared with the preset benchmark value to identify the risk of trading behavior in the current stage. The risk features that deviate from the benchmark value of the stage are identified, which means that the risk of trading behavior in the current stage is determined;

[0032] In the early stages of a transaction, the input risk characteristics, including the counterparty's credit score, historical transaction default rate, and supply and demand change rate, are compared with the preset benchmark values of the risk characteristics to determine the risk characteristics that deviate from the benchmark values, and then calculate the risk trend index.

[0033] During the mid-term of a transaction, the input mid-term risk characteristics, including price volatility, cargo transportation delay rate, and number of days of delayed funds arrival, are compared with the preset mid-term risk characteristic benchmark values to determine the mid-term risk characteristics that deviate from the benchmark values, and then calculate the mid-term risk trend index;

[0034] At the later stage of a transaction, the inputted later stage risk characteristics, including the number of transaction disputes, product delivery quality qualification rate, and accounts receivable collection period, are compared with the preset benchmark values of the later stage risk characteristics to determine the later stage risk characteristics that deviate from the benchmark values, and then calculate the later stage risk trend index;

[0035] After the risk identification model completes the analysis, it obtains the risk trend analysis results of the transaction behavior in the current stage, and then records the risk trend analysis results in the system's risk identification log, which is associated with the detailed information of the transaction behavior.

[0036] A further improvement of the technical solution of the present invention is that the calculation process of the early stage risk trend index is:

[0037] Extracting prior risk characteristic data, including counterparty credit scores, historical transaction default rates, and supply and demand change rates, and obtaining preset counterparty credit score benchmark values, historical transaction default rate benchmark values, and supply and demand change rate benchmark values, and calculating the difference ratio between each prior risk characteristic and its benchmark value;

[0038] Take the square of each calculated difference ratio, add up the square results to get a sum, take the square root of the sum, and multiply it by 1 / 3 to get the previous period risk trend index;

[0039] The calculation process of the medium-term risk trend index is as follows:

[0040] Extract medium-term risk characteristic data, including price volatility, freight transportation delay rate, and number of days of delayed funds arrival. Obtain preset benchmark values for price volatility, freight transportation delay rate, and number of days of delayed funds arrival, and calculate the logarithmic ratio of each medium-term risk characteristic to its benchmark value.

[0041] Add up the calculated logarithmic ratios of each value to get a sum, divide the sum by 3 to get the average value, and then take the negative exponential function value of the average value to finally get the medium-term risk trend index;

[0042] The calculation process of the late risk trend index is:

[0043] Extracting late-stage risk feature data, including the number of transaction disputes, product delivery quality pass rate, and accounts receivable collection cycle, and obtaining preset benchmark values for the number of transaction disputes, product delivery quality pass rate, and accounts receivable collection cycle, and calculating the ratio of each late-stage risk feature to its benchmark value;

[0044] Each calculated ratio is squared separately, and all square results are added together to obtain a sum, and then the cube root of the sum is taken to finally obtain the later risk trend index.

[0045] A further improvement of the technical solution of the present invention is that the risk warning module specifically includes:

[0046] Obtain the risk identification and risk trend analysis results of transaction behaviors from the risk trend analysis module, and preliminarily organize the identification and analysis results to extract the early-stage risk trend index, mid-term risk trend index, and late-stage risk trend index;

[0047] Based on the risk trend index obtained at each stage, the trading behavior is assessed for risk, and corresponding risk warning thresholds are set for the early, mid, and late stages of the trading behavior. The risk warning thresholds are set based on historical data, industry experience, and user demand factors;

[0048] Real-time monitoring of the comparison between the risk trend index and the set risk warning threshold. When the risk trend index exceeds the preset threshold of the corresponding stage, the warning signal is triggered, and risk warning information containing detailed information of transaction behavior, description of risk characteristics and warning suggestions is generated. The risk warning information is sent to relevant users through the system's preset notification channels, such as SMS, email or system messages, so that users can take quick measures to deal with potential risks and reduce possible losses.

[0049] A further improvement of the technical solution of the present invention is that the countermeasure recommendation module specifically includes:

[0050] The response measure recommendation module receives risk warning information from the risk warning module and analyzes the warning information. At the same time, based on the risk characteristics and stage in the warning information, it preliminarily determines the direction of recommended response measures;

[0051] Combining the transaction-related party information and industry knowledge in the knowledge graph, we analyze the background, historical transaction records, and market conditions of both parties to the transaction, identify the root causes and scope of impact of the risk, and select response measures that match the current risk characteristics and the current stage from the pre-established response strategy library. We match appropriate response measures for different risks and transaction stages, and then present the recommended results to users in a clear and easy-to-understand manner.

[0052] A further improvement of the technical solution of the present invention is that the risk assessment report generation module specifically includes:

[0053] The risk assessment report generation module collects data from various system modules according to a set cycle, obtains risk assessment results before, during, and after the transaction, obtains risk trend analysis data, organizes risk warning data, records warning time, content, and trends, and then summarizes response measures and implementation status;

[0054] Automatically generate a cross-border bulk commodity e-commerce transaction risk assessment report based on the collated risk warning data and the preset report template and format;

[0055] The generated risk assessment report will be distributed through the system's preset output channels such as emails and system messages to ensure that relevant users can obtain the report content in a timely manner. At the same time, the risk assessment report will be stored in the system's database to facilitate users' subsequent review and historical data comparison and analysis, and the risk assessment report will be archived in chronological order to establish a complete risk assessment file.

[0056] Due to the adoption of the above technical solution, the present invention has the following technical advancements compared to the prior art:

[0057] 1. The present invention provides a risk assessment and management system for cross-border bulk commodity e-commerce transactions. By constructing a risk identification model and combining the risk knowledge and association relationships in the knowledge graph, risks are identified before, during, and after the transaction. At the same time, corresponding risk warning thresholds are set for different periods of the transaction to trigger warning signals, enabling enterprises to take measures before or in the early stages of the risk, effectively avoiding or reducing losses caused by the risk, and improving the enterprise's ability to respond to risks.

[0058] 2. The present invention provides a cross-border bulk commodity e-commerce transaction risk assessment and management system, which combines a pre-established response strategy library, based on the warning information issued by the risk warning module, combined with the transaction-related party information and industry knowledge in the knowledge graph, and intelligently recommends appropriate response measures from the response strategy library. It can match appropriate response strategies according to different risks and transaction stages, provide reasonable decision-making support for enterprises, improve the efficiency and accuracy of enterprises in responding to risks, and reduce decision-making costs.

[0059] 3. The present invention provides a cross-border bulk commodity e-commerce transaction risk assessment and management system, which controls the entire process of cross-border bulk commodity e-commerce transaction risk assessment, from pre-transaction risk assessment, to real-time monitoring during the transaction, and then to post-transaction summary analysis. Each link works closely together to ensure the comprehensiveness and continuity of risk management. The full-process control helps enterprises to promptly discover and respond to potential risks, reduce risk losses, and improve the security and stability of transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0061] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0062] Figure 2 Schematic diagram of the system function modules of the present invention. DETAILED DESCRIPTION

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0064] Example 1, as Figure 1 、 Figure 2 As shown, the present invention provides a cross-border bulk commodity e-commerce transaction risk assessment and management system, including a transaction risk management center, which is communicatively connected to a data collection and integration module, a knowledge graph module, a risk trend analysis module, a risk warning module, a response measure recommendation module, and a risk assessment report generation module, wherein the modules are electrically connected;

[0065] The transaction risk management center is responsible for coordinating the communication and data interaction between various modules. It can receive the basic transaction data set from the data collection and integration module, send instructions to the knowledge graph module to update the knowledge graph content, receive the risk identification and trend analysis results of the risk trend analysis module, and control the risk warning module to issue warning signals. It also integrates the recommendations of the response measure recommendation module and the content of the risk assessment report generation module, realizing the full process control of cross-border bulk commodity e-commerce transaction risk assessment, ensuring close cooperation among all links, and improving the coordination and efficiency of the entire system operation.

[0066] The data collection and integration module is used to collect data related to cross-border bulk commodity transactions from multiple data sources, including but not limited to order information, payment information, logistics information, supplier and buyer credit data, market price fluctuation data, etc. The data sources include cross-border e-commerce platforms, third-party payment platforms, logistics companies, external market data, etc., and pre-process and integrate the collected data to obtain a basic transaction data set, integrating multiple interfaces, and collecting data related to cross-border bulk commodity transactions from multiple data sources including cross-border e-commerce platforms, third-party payment platforms, logistics companies, external market data, etc., including but not limited to order information, payment information, logistics information, supplier and buyer credit data, market price fluctuation data, etc. Among them, order information is captured from cross-border e-commerce platforms, covering product details, transaction time, quantity, etc., and payment information is obtained from third-party payment platforms, including payment method, amount, status, etc., and is integrated with logistics companies. The system is connected to the company to collect logistics information, including transportation tracks, receipt status, etc. At the same time, the credit data of suppliers and buyers, and market price fluctuation data are obtained from external market data sources and internal credit assessment systems to ensure that the data fully covers all aspects of cross-border bulk commodity transactions. The collected data related to cross-border bulk commodity transactions are pre-processed, including data cleaning, outlier processing and standardization. Through data cleaning, duplicate, erroneous and missing value records are removed. Standardization processing is performed on data in different formats to unify the data format, outliers are identified and corrected, and data that deviates significantly from the market price is corrected. The different types of pre-processed data are associated and integrated according to the transaction order number field, and the order information, payment information, logistics information, credit data and market price fluctuation data are integrated to form a complete and coherent transaction record. The basic transaction data set is constructed through data mapping and conversion;

[0067] The knowledge graph module is used to build a knowledge graph of transaction behavior, integrate various types of information involved in the transaction, including suppliers, buyers, logistics providers, market conditions and historical transaction records, screen out risk features related to transaction risk assessment, and form a risk feature sequence list. It obtains the basic transaction data set from the data acquisition and integration module, uses natural language processing technology to parse the text data involved, combines with the rule engine, and identifies and extracts entities in the basic transaction data set according to the preset rule template, including supplier name, buyer information, logistics order number, etc., and establishes the relationship between entities, integrating scattered data points into a structured information network. At the same time, it extracts the attribute information in the data and constructs a preliminary knowledge graph structure. Based on the actual needs of cross-border commodity transaction risk assessment, risk characteristics related to transaction risk assessment are screened out from the integrated information, and transaction behaviors are divided into three periods: early stage, mid-term and late stage. The key risk characteristics of each stage of the early, mid-term and late stage of transaction behaviors are clarified. Among them, the early risk characteristics of transaction behaviors include counterparty credit score, historical transaction default rate and supply and demand change rate. The mid-term risk characteristics of transaction behaviors include price volatility, cargo transportation delay rate and number of days of delay in funds arrival. The late risk characteristics of transaction behaviors include the number of transaction disputes, product delivery quality qualification rate and accounts receivable collection cycle. The counterparty credit score can intuitively reflect the credit reliability of the counterparty. The lower the score, the greater the possibility of counterparty default. The higher the transaction risk, the higher the historical transaction default rate is. The ratio of the number of transactions in which default occurred to the total number of transactions when the counterparty participated in cross-border bulk commodity e-commerce transactions in the past year. The supply and demand change rate is the change in the market supply and demand of the target bulk commodity in the two months before the transaction. Supply and demand changes directly affect the price trend of bulk commodities. If the supply and demand change rate is large, it means that the market is in an unstable state and the risk of price fluctuation increases, which in turn affects the profit expectations and risk level of the transaction. The price volatility rate is the fluctuation range of the target bulk commodity price in two adjacent days during the transaction. The price volatility rate reflects the degree of market price fluctuation. The higher the price volatility rate, the more unstable the market price is, and the trading parties may face greater price risks. The cargo transportation delay rate is the ratio of the number of transactions in which the actual transportation time of goods exceeds the expected transportation time to the total number of transactions during the transaction process. Cargo transportation delay will affect the delivery period of the transaction. The higher the delay rate, the greater the risk in the logistics link. The number of days of delay in funds arrival is calculated from the agreed payment time of the transaction, and the difference between the actual funds arrival time and the agreed time. The delay in funds arrival will affect the smooth progress of the transaction. The more days of delay, the greater the risk of capital flow, which may have an adverse impact on the transaction. The number of transaction disputes is the number of disputes between the buyer and the seller due to issues such as product quality, quantity, and delivery time after the completion of the transaction. A higher number of disputes indicates that there may be problems with the transaction, which will affect the subsequent cooperation between the two parties and the reputation of the company.The product delivery quality pass rate measures whether the quality of the delivered commodities meets the contractually agreed standards. It calculates the ratio of the number of qualified products to the total number of delivered products. A low pass rate means increased costs and losses for the company. The accounts receivable collection cycle, measured in days, measures the time from transaction completion to the actual recovery of all accounts receivable. A longer collection cycle increases capital utilization costs and bad debt risk, impacting the company's capital turnover and financial status. The screened risk features are categorized and organized according to the early, mid, and late stages of the transaction to form a risk feature sequence table. The risk feature sequence table clearly lists the risk features of each stage, assigns corresponding identifiers and attribute descriptions to each risk feature, and maps the risk feature sequence table to the knowledge graph to form a subgraph structure containing risk feature nodes and association relationships.

[0068] The risk trend analysis module is used to build a risk identification model based on the risk characteristics in the risk characteristic sequence table, identify risks at the beginning, middle, and end of the transaction behavior, and analyze the risk trend of the transaction behavior period based on the deviation of each risk characteristic from its preset benchmark value. The risk trend analysis module includes a risk identification model construction unit and a risk identification unit;

[0069] Among them, the risk identification model construction unit is used to utilize the historical transaction data stored in the knowledge graph and combine the risk features related to transaction risk assessment in the risk feature sequence table to construct a risk identification model based on the neural network model, extract historical transaction data related to transaction risk assessment from the knowledge graph, and screen out risk feature data related to transaction risk assessment based on the risk feature types covered by the risk feature sequence table, normalize the features, scale the numerical features to the range of 0 to 1, and eliminate dimensional differences. At the same time, combined with user needs and historical transaction data, preset the benchmark value of each risk feature, mark the risk features that deviate from the benchmark value, and integrate them into a comprehensive feature data set. According to the characteristics and needs of cross-border commodity transaction risk assessment, a neural network architecture is selected to build a multi-layer perceptron (MLP) network model, which includes an input layer, multiple hidden layers and an output layer. The number of neurons in the input layer matches the number of extracted risk features, and each neuron corresponds to a risk feature. The processed feature data is input into the model. The hidden layer consists of several neurons, which are activated by The active function introduces nonlinear factors to enable the model to learn complex risk feature combinations and relationships. The appropriate number of hidden layer neurons is set to increase the model's expressiveness and flexibility. The connection weights and biases of the hidden layer are adjusted. During the training process, the model is updated using the backpropagation algorithm to adapt to the risk identification task and output risk features that deviate from the baseline value. The prepared feature data set is divided into a training set and a validation set. The constructed neural network model is trained using the training set. The model training process is adjusted by setting hyperparameters such as the learning rate and the number of iterations. During the training process, the loss function is calculated using the backpropagation algorithm, and the network weights are adjusted to minimize the loss function value so that the model learns the mapping relationship between risk features and transaction risks. During the training process, the model is verified using the validation set to evaluate the model's performance on unseen data and prevent model overfitting. By adjusting hyperparameters and optimizing the network structure, the model achieves good performance on the validation set, enabling the model to accurately identify risks in cross-border commodity transactions, until the model achieves satisfactory performance and obtains the final risk identification model.

[0070] The risk identification unit is used to use the constructed risk identification model to identify risks in different periods of transaction behavior, analyze the deviation of each risk feature from its own benchmark value, calculate the early risk trend index, the mid-term risk trend index and the late risk trend index, clarify the risk trend of each period of transaction behavior, input the pre-processed risk feature data into the constructed risk identification model, compare it with the preset benchmark value, and then identify the risk of transaction behavior in the current stage, identify the risk features that deviate from the benchmark value of the stage, that is, determine the risk of transaction behavior in the stage. In the early stage of the transaction behavior, based on the input early risk features including the counterparty credit score, historical transaction default rate and supply and demand change rate, and compare them with the preset benchmark value of the early risk features, the early risk deviation from the benchmark value is determined. Characteristics, and then calculate the early risk trend index. In the middle of the transaction behavior, the input medium-term risk characteristics including price volatility, cargo transportation delay rate and number of days of delayed funds arrival are compared with the preset benchmark value of the medium-term risk characteristics to determine the medium-term risk characteristics that deviate from the benchmark value, and then calculate the medium-term risk trend index. In the late stage of the transaction behavior, the input late stage risk characteristics including the number of transaction disputes, product delivery quality qualification rate and accounts receivable collection period are compared with the preset benchmark value of the late stage risk characteristics to determine the late stage risk characteristics that deviate from the benchmark value, and then calculate the late stage risk trend index. After the risk identification model completes the analysis, it obtains the risk trend analysis results of the transaction behavior at the current stage, and then records the risk trend analysis results in the risk identification log of the system and associates them with the detailed information of the transaction behavior;

[0071] In addition, the calculation process of the previous risk trend index is:

[0072] Extracting prior risk characteristic data, including counterparty credit scores, historical transaction default rates, and supply and demand change rates, and obtaining preset counterparty credit score benchmark values, historical transaction default rate benchmark values, and supply and demand change rate benchmark values, calculating the difference ratio between each prior risk characteristic and its benchmark value, squaring each calculated difference ratio, and then adding the squared results to obtain a sum. Taking the square root of the sum and multiplying it by 1 / 3, ultimately obtaining a prior risk trend index;

[0073] The expression of the previous risk trend index is:

[0074]

[0075] Where R Q is the previous risk trend index, C s is the current counterparty credit score, C b，s is the preset counterparty credit score benchmark value, H d is the current historical transaction default rate, H b，dis the preset historical transaction default rate benchmark value, S r is the current supply and demand change rate, S b，r is the preset supply and demand change rate benchmark value, when C s The closer to C b，s 、H d The closer to H b，d 、S r The closer to S b，r When R Q Approaching 0, it means that the risk trend of the current transaction in the early stage is similar to the benchmark situation, and the risk is relatively stable. s 、H d 、S r The greater the deviation from the corresponding reference value, the greater the Q An increase indicates that there is a significant change in the previous risk trend, and the risk level deviating from the baseline situation is higher;

[0076] The calculation process of the medium-term risk trend index is:

[0077] Extract medium-term risk characteristic data including price volatility, freight transportation delay rate, and number of days of delayed funds arrival, obtain preset benchmark values for price volatility, freight transportation delay rate, and number of days of delayed funds arrival, calculate the logarithmic ratio of each medium-term risk characteristic to its benchmark value, add up each calculated logarithmic ratio to obtain a sum, divide the sum by 3 to obtain an average, then apply the negative exponential function value to the average to finally obtain the medium-term risk trend index;

[0078] The expression of medium-term risk trend index is:

[0079]

[0080] Where R Z is the medium-term risk trend index, P v is the current price volatility, P b，v is the preset price volatility benchmark value, G d is the current cargo transportation delay rate, G b，d is the preset cargo transportation delay rate benchmark value, F a F is the current delay in the arrival of funds. b，a is the preset benchmark value of the number of days of delay in the arrival of funds. v >P b，v , G d >G b，d 、F a >F b，a When the price volatility, cargo transportation delay rate and funds arrival delay days are all higher than the benchmark value, R Z A decrease indicates that the medium-term risk trend deviates significantly from the baseline and the risk level increases;

[0081] The calculation process of the late risk trend index is:

[0082] Extract late-stage risk characteristic data, including the number of transaction disputes, product delivery quality pass rate, and accounts receivable collection cycle. Obtain preset benchmark values for the number of transaction disputes, product delivery quality pass rate, and accounts receivable collection cycle. Calculate the ratio of each late-stage risk characteristic to its benchmark value. Square each calculated ratio, add all squared results, and obtain a sum. Take the cube root of this sum to ultimately obtain a late-stage risk trend index.

[0083] The expression of the late risk trend index is:

[0084]

[0085] Where R H is the later risk trend index, T d is the current number of transaction disputes, T b，d is the preset benchmark value of transaction disputes, Q q is the current product delivery quality qualification rate, Q b，q A is the preset product delivery quality qualification rate benchmark value, c is the current accounts receivable collection period, A b，c is the preset benchmark value of accounts receivable collection period. d The closer to T b，d , Q q The closer to Q b，q 、A c The closer to A b，c When R H Approaching 0, it means that the risk trend of the current transaction in the later period is consistent with the benchmark situation, and the risk is relatively stable. d , Q q 、A c The greater the deviation from the corresponding reference value, the greater the H Increased, indicating that the risk trend has changed significantly in the later period, and the risk level is higher;

[0086] The risk warning module is used to conduct risk assessments on trading behaviors based on the results of risk identification and risk trend analysis. It also sets corresponding risk warning thresholds for different periods of trading behaviors to trigger warning signals and issue risk warning information.

[0087] The response measure recommendation module is used to combine the pre-established response strategy library containing various risk response strategies, based on the risk warning information issued by the risk warning module, and combined with the transaction-related party information and industry knowledge in the knowledge graph, to intelligently recommend appropriate response measures from the response strategy library;

[0088] The risk assessment report generation module is used to automatically generate cross-border bulk commodity e-commerce transaction risk assessment reports according to the set cycle, including risk assessment results before, during and after the transaction, risk trend analysis, risk warning status, response measures recommendations and implementation status.

[0089] Example 2, as Figure 1 、 Figure 2 As shown, based on Example 1, the present invention provides a technical solution: preferably, the risk warning module specifically includes:

[0090] Obtain the risk identification results and risk trend analysis results of the transaction behavior from the risk trend analysis module, and preliminarily organize the identification and analysis results, extract the early stage risk trend index, mid-term risk trend index and late stage risk trend index, conduct risk assessment on the transaction behavior based on the risk trend index of each stage obtained, and set corresponding risk warning thresholds for the early stage, mid-term and late stages of the transaction behavior respectively. The risk warning threshold is set based on historical data, industry experience and user demand factors, and the comparison between the risk trend index and the set risk warning threshold is monitored in real time. When the risk trend index exceeds the preset threshold of the corresponding stage, a warning signal is triggered, and risk warning information containing detailed information of the transaction behavior, risk feature description and warning suggestions is generated. The risk warning information is sent to relevant users through the system's preset notification channels, such as SMS, email or system messages, so that users can quickly take measures to deal with potential risks and reduce possible losses;

[0091] The recommended response module specifically includes:

[0092] The countermeasure recommendation module receives risk warning information from the risk warning module and analyzes the warning information. At the same time, based on the risk characteristics and stage in the warning information, it preliminarily determines the direction of recommended countermeasures. Combined with the transaction-related party information and industry knowledge in the knowledge graph, it analyzes the background, historical transaction records and market conditions of both parties to the transaction, identifies the root cause and scope of the risk, and selects countermeasures that match the current risk characteristics and stage from the pre-established countermeasure strategy library. It matches appropriate countermeasures for different risks and transaction stages, and then presents the recommended results to users in a clear and understandable manner. For example, for counterparties with high credit risk, it recommends providing guarantees or early collection of payments. For price volatility risks, it recommends hedging or adjusting transaction terms.

[0093] The risk assessment report generation module specifically includes:

[0094] The risk assessment report generation module collects data from various modules of the system according to the set cycle, obtains the risk assessment results before, during and after the transaction, obtains risk trend analysis data, organizes risk warning data, records the warning time, content and trend, and then summarizes the response measures recommended and the implementation status. According to the organized risk warning data, according to the preset report template and format, it automatically generates a cross-border bulk commodity e-commerce transaction risk assessment report. The report content presents the risk status before, during and after the transaction in detail, describes the risk trend analysis results through charts and text, records the risk warning situation, including warning time, risk characteristics and recommended measures, shows the recommendation of response measures and the evaluation of implementation effect, and puts forward improvement suggestions. The generated risk assessment report is distributed through the system's preset output channels such as emails and system messages to ensure that relevant users can obtain the report content in a timely manner. At the same time, the risk assessment report is stored in the system's database to facilitate users' subsequent review and historical data comparison and analysis, and archives the risk assessment report in chronological order to establish a complete risk assessment file.

[0095] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A cross-border bulk commodity e-commerce transaction risk assessment and management system, including a transaction risk management center, characterized by: The transaction risk management center is communicatively connected to a data collection and integration module, a knowledge graph module, a risk trend analysis module, a risk warning module, a response measure recommendation module, and a risk assessment report generation module, wherein the modules are electrically connected; The data collection and integration module is used to collect data related to cross-border commodity transactions from multiple data sources, and pre-process and integrate the collected data to obtain a basic transaction data set; The knowledge graph module is used to construct a knowledge graph of transaction behavior, screen out risk features related to transaction risk assessment, and form a risk feature sequence list; The risk trend analysis module is used to construct a risk identification model based on the risk features in the risk feature sequence table, identify risks at the early, middle, and late stages of a transaction, and analyze the risk trend of the transaction period based on the deviation of each risk feature from its own preset benchmark value; The risk warning module is used to conduct risk assessment of transaction behaviors based on the results of risk identification and risk trend analysis, trigger warning signals, and issue risk warning information; The response measure recommendation module is used to intelligently recommend appropriate response measures from the response strategy library based on the risk warning information in combination with the pre-established response strategy library; The risk assessment report generation module is used to automatically generate a cross-border bulk commodity e-commerce transaction risk assessment report.

2. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 1, characterized in that: The data acquisition and integration module specifically includes: Integrate multiple interfaces to collect data related to cross-border commodity transactions from multiple data sources, including cross-border e-commerce platforms, third-party payment platforms, logistics companies, and external market data. This includes capturing order information from cross-border e-commerce platforms, obtaining payment information from third-party payment platforms, connecting with logistics company systems to collect logistics information, and obtaining supplier and buyer credit data and market price fluctuation data from external market data sources and internal credit assessment systems. Pre-processing of collected data related to cross-border commodity transactions, including data cleaning, outlier processing, and standardization; The different types of pre-processed data are associated and integrated according to the fields of the transaction order number, and the order information, payment information, logistics information, credit data and market price fluctuation data are integrated to form a complete and coherent transaction record. Through data mapping and conversion, the basic transaction data set is constructed.

3. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 1, characterized in that: The knowledge graph module specifically includes: The basic transaction data set is obtained from the data collection and integration module. Natural language processing technology is used to parse the text data involved. In combination with the rule engine, based on the preset rule template, the entities in the basic transaction data set are identified and extracted. The relationship between the entities is established, and the scattered data points are integrated into a structured information network. At the same time, the attribute information in the data is extracted to construct a preliminary knowledge graph structure. Based on the actual needs of cross-border commodity trading risk assessment, risk characteristics related to transaction risk assessment are screened from the integrated information, and trading behavior is divided into three periods: early stage, mid-term, and late stage. The key risk characteristics of each stage of trading behavior are clearly defined. Among them, early stage risk characteristics of trading behavior include counterparty credit score, historical transaction default rate, and supply and demand change rate. Mid-term risk characteristics of trading behavior include price volatility, cargo transportation delay rate, and number of days of delay in funds arrival. Late stage risk characteristics of trading behavior include the number of transaction disputes, product delivery quality qualification rate, and accounts receivable collection period. The screened risk features are classified and organized according to the early, middle and late stages of the transaction behavior to form a risk feature sequence list. Each risk feature is assigned a corresponding identifier and attribute description, and the risk feature sequence list is mapped to the knowledge graph to form a subgraph structure containing risk feature nodes and association relationships.

4. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 3, characterized in that: The risk trend analysis module includes a risk identification model building unit and a risk identification unit; The risk identification model construction unit is used to utilize the historical transaction data stored in the knowledge graph and, in combination with the risk features related to transaction risk assessment in the risk feature sequence table, to construct a risk identification model based on a neural network model; The risk identification unit is used to use the constructed risk identification model to identify risks in different periods of trading behavior, analyze the deviation of each risk characteristic from its respective benchmark value, calculate the early risk trend index, the mid-term risk trend index and the late risk trend index, and clarify the risk trend of each period of trading behavior.

5. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 4, characterized in that: The risk identification model building unit specifically includes: Extract historical transaction data related to transaction risk assessment from the knowledge graph, and screen out risk feature data related to transaction risk assessment based on the risk feature types covered by the risk feature sequence table. Normalize the features, and, based on user needs and historical transaction data, preset baseline values for each risk feature. Mark risk features that deviate from the baseline values, and integrate them into a comprehensive feature dataset. Based on the characteristics and requirements of cross-border commodity trading risk assessment, a neural network architecture was selected to build a multi-layer perceptron network model, consisting of an input layer, multiple hidden layers, and an output layer. The prepared feature data set is divided into a training set and a validation set. The training set is used to train the constructed neural network model, and the validation set is used to validate the model to obtain the final risk identification model.

6. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 4, characterized in that: The risk identification unit specifically includes: The pre-processed risk feature data is input into the established risk identification model and compared with the preset benchmark value to identify the risk of trading behavior in the current stage. The risk features that deviate from the benchmark value of the stage are identified, which means that the risk of trading behavior in the current stage is determined; In the early stages of a transaction, the input risk characteristics, including the counterparty's credit score, historical transaction default rate, and supply and demand change rate, are compared with the preset benchmark values of the risk characteristics to determine the risk characteristics that deviate from the benchmark values, and then calculate the risk trend index. During the mid-term of a transaction, the input mid-term risk characteristics, including price volatility, cargo transportation delay rate, and number of days of delayed funds arrival, are compared with the preset mid-term risk characteristic benchmark values to determine the mid-term risk characteristics that deviate from the benchmark values, and then calculate the mid-term risk trend index; At the later stage of a transaction, the inputted later stage risk characteristics, including the number of transaction disputes, product delivery quality qualification rate, and accounts receivable collection period, are compared with the preset benchmark values of the later stage risk characteristics to determine the later stage risk characteristics that deviate from the benchmark values, and then calculate the later stage risk trend index; After the risk identification model completes the analysis, it obtains the risk trend analysis results of the transaction behavior in the current stage, and then records the risk trend analysis results in the system's risk identification log, which is associated with the detailed information of the transaction behavior.

7. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 6, characterized in that: The calculation process of the previous risk trend index is as follows: Extracting prior risk characteristic data, including counterparty credit scores, historical transaction default rates, and supply and demand change rates, and obtaining preset counterparty credit score benchmark values, historical transaction default rate benchmark values, and supply and demand change rate benchmark values, and calculating the difference ratio between each prior risk characteristic and its benchmark value; Take the square of each calculated difference ratio, add up the square results to get a sum, take the square root of the sum, and multiply it by 1 / 3 to finally get the previous period risk trend index; The calculation process of the medium-term risk trend index is as follows: Extract medium-term risk characteristic data, including price volatility, freight transportation delay rate, and number of days of delayed funds arrival. Obtain preset benchmark values for price volatility, freight transportation delay rate, and number of days of delayed funds arrival, and calculate the logarithmic ratio of each medium-term risk characteristic to its benchmark value. Add up the calculated logarithmic ratios of each value to get a sum, divide the sum by 3 to get the average value, and then take the negative exponential function value of the average value to finally get the medium-term risk trend index; The calculation process of the late risk trend index is: Extracting late-stage risk feature data, including the number of transaction disputes, product delivery quality pass rate, and accounts receivable collection cycle, and obtaining preset benchmark values for the number of transaction disputes, product delivery quality pass rate, and accounts receivable collection cycle, and calculating the ratio of each late-stage risk feature to its benchmark value; Each calculated ratio is squared separately, and all square results are added together to obtain a sum, and then the cube root of the sum is taken to finally obtain the later risk trend index.

8. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 4, characterized in that: The risk warning module specifically includes: Obtain the risk identification and risk trend analysis results of transaction behaviors from the risk trend analysis module, and preliminarily organize the identification and analysis results to extract the early-stage risk trend index, mid-term risk trend index, and late-stage risk trend index; Conduct risk assessments on trading behaviors based on the risk trend indexes obtained at each stage, and set corresponding risk warning thresholds for the early, mid, and late stages of trading behaviors; Real-time monitoring of the comparison between the risk trend index and the set risk warning threshold. When the risk trend index exceeds the preset threshold of the corresponding stage, the warning signal is triggered, and risk warning information containing detailed information of transaction behavior, risk feature description and warning suggestions is generated. The risk warning information is sent to relevant users through the notification channels preset by the system.

9. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 8, characterized in that: The countermeasure recommendation module specifically includes: The response measure recommendation module receives risk warning information from the risk warning module and analyzes the warning information. At the same time, based on the risk characteristics and stage in the warning information, it preliminarily determines the direction of recommended response measures; Combining the transaction-related party information and industry knowledge in the knowledge graph, the background, historical transaction records and market conditions of the two parties to the transaction are analyzed to clarify the root cause and scope of the risk. Then, from the pre-established response strategy library, response measures that match the current risk characteristics and the current stage are selected. For different risks and transaction stages, appropriate response measures are matched, and the recommended results are presented to the user.

10. The cross-border bulk commodity e-commerce transaction risk assessment and management system according to claim 1, characterized in that: The risk assessment report generation module specifically includes: The risk assessment report generation module collects data from various system modules according to the set cycle, obtains risk assessment results before, during, and after the transaction, obtains risk trend analysis data, organizes risk warning data, and then summarizes the response measures recommended and their implementation status; Automatically generate a cross-border bulk commodity e-commerce transaction risk assessment report based on the collated risk warning data and the preset report template and format; The generated risk assessment report will be distributed through the system's preset email and system message output channels. At the same time, the risk assessment report will be stored in the system's database and archived in chronological order to establish a complete risk assessment file.

Images