A Method and System for Digital Optimization of Trade Processes

Abstract

The present invention relates to the technical field of trade operation, and specifically provides a method and system for digital optimization of trade processes, including the following steps: extracting the time-series characteristic values, logistics node lag time values, and abnormal transaction behavior amounts of transaction orders from a trade data set, and extracting the range of transaction behavior characteristics through characteristic difference operations. In the present invention, by deeply analyzing transaction data, extracting the time-series characteristics of transaction orders and the lag time of logistics nodes, accurately screening transaction behavior characteristics, optimizing data association, improving the coordination of the transaction chain, dynamically identifying changes in key nodes based on global data trend analysis, ensuring time-series coherence and optimizing the connectivity between nodes, enhancing the collaborative ability between nodes through fine-grained time-series analysis, reducing the impact of abnormal delays, constructing a dynamic weight matrix, real-time monitoring and identifying potential risks, adjusting path flow and weight distribution, realizing the optimal allocation of resources, and reducing supply chain pressure.

Description

Technical Field

[0001] The present invention relates to the technical field of trade operations, and in particular to a method and system for digital optimization of trade processes. Background Art

[0002] The technical field of trade operations includes research on the management and optimization of trade-related processes, activities, and resources. Its core contents include data processing and technical implementation in links such as commodity trading, logistics management, supply chain collaboration, order processing, payment settlement, etc. This technical field improves the efficiency of trade activities through digital means, with the focus on establishing an efficient and secure trading system, covering transaction behavior records, commodity and service information management, transaction data statistical analysis, and the full-process coordination of the supply chain, aiming to achieve comprehensive digital management and efficient utilization of trade resources.

[0003] Among them, the method for digital optimization of trade processes refers to improving and optimizing trade processes through digital technical means, covering the classification and sorting of transaction information, the digital conversion of trade resources, the automated operation of transaction processes, and the data docking and processing of trade nodes, etc. This method realizes real-time monitoring and adjustment of key steps in the transaction process by collecting, processing, transmitting, and integrating information in the trade process, and combines specific data sorting methods to ensure the consistency and standardization of trade operation processes.

[0004] Existing technologies rely mostly on fixed rules in data processing and are unable to flexibly respond to dynamic changes in transaction processes, resulting in weak adaptability of the transaction chain and difficulty in timely adjustment. The recognition accuracy of key transaction nodes is relatively low, lacking in-depth mining of the timing characteristics of nodes, which affects the coordination of the overall process. The adjustment of node connection weights lags behind, unable to make full use of the changing trend of transaction data, leading to unbalanced resource allocation and affecting logistics efficiency. There is a lack of effective means for monitoring risk conduction, which easily causes local risks to spread to the entire trade process, increasing the uncertainty of the business. During the path optimization process, there is a lack of precise adjustment of the dynamic distribution of traffic, affecting the overall smoothness of the transaction chain. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose a method and system for digital optimization of trade processes.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: A method for digital optimization of trade processes, including the following steps:

[0007] S1: Extract the time series eigenvalue of the transaction order, the lag time value of the logistics node, and the abnormal transaction behavior volume from the trade data set. Extract the range of transaction behavior features through feature difference operation, compare the difference values between the data, adjust the feature screening rules to optimize the correlation features of the transaction behavior and the logistics node, and generate a sparse feature vector set;

[0008] S2: Based on the sparse feature vector set, calculate the global trade data change trend sequence value, extract the dynamic trend of the core transaction node, judge whether the lag time of the key node meets the standard, analyze the change of the connectivity weight of the core transaction node, verify the consistency, and establish the global key node trend prediction value;

[0009] S3: Based on the global key node trend prediction value, separate the fine-grained node time change data, calculate the time series correlation metric value of each node, extract the time change rate value, compare the connectivity weight values of adjacent nodes, call the node update rule to perform correlation iterative optimization, and establish a distributed key node dynamic weight matrix;

[0010] S4: Based on the distributed key node dynamic weight matrix, extract the combined value of the risk factors between nodes, calculate the risk conduction probability value, analyze the abnormal change area of the weight matrix, screen the nodes with significant risks, compare the dynamic weight differences between nodes, and establish a node correlation risk transmission index;

[0011] S5: Based on the node correlation risk transmission index, construct an optimized path sequence of the risk node, adjust the flow value and weight value of the critical path, verify the connectivity of the optimized path, adjust the flow distribution ratio, and generate a set of optimized paths for the key nodes.

[0012] As a further solution of the present invention, the sparse feature vector set includes the range of transaction behavior features, the optimized value of the feature screening rules, and the correlation features of the transaction behavior and the logistics node. The global key node trend prediction value includes the dynamic change trend of the core transaction node, the standard value of the lag time of the key node, and the change of the connectivity weight of the core transaction node. The distributed key node dynamic weight matrix includes the fine-grained node time change data, the time series correlation metric value, the time change rate value, and the connectivity weight values of adjacent nodes. The node correlation risk transmission index includes the combined value of the risk factors between nodes, the risk conduction probability value, the abnormal change area of the weight matrix, the nodes with significant risks, and the dynamic weight differences between nodes. The set of optimized paths for the key nodes includes the optimized path sequence of the risk node, the flow value of the critical path, the weight value, and the flow distribution ratio.

[0013] As a further solution of the present invention, the specific steps for obtaining the sparse feature vector set are as follows:

[0014] S101: Extract the time-series eigenvalue of the transaction order from the trade data set, sort the extracted time-series eigenvalue, generate a time difference sequence by calculating the time difference between transaction orders, filter the transaction orders in the high-frequency trading time period according to the distribution characteristics of the time difference sequence, calculate the intersection ratio between the high-frequency trading time period and the overall trading time series, and generate a trading high-frequency feature interval;

[0015] S102: Cross-compare the time period eigenvalue in the trading high-frequency feature interval with the logistics node lag time value, calculate the difference mean and difference standard deviation of each period according to the interval, filter the intervals where the difference mean is greater than a certain multiple of the standard deviation, and generate a high-difference feature interval set by comparing the interval coverage rate;

[0016] S103: Cumulatively calculate the abnormal trading behavior volume in the high-difference feature interval set, using the formula:

[0017] ;

[0018] Calculate the abnormal trading behavior impact value of multiple intervals, and generate an abnormal trading behavior impact value set;

[0019] Among them, represents the abnormal trading behavior impact value, represents the time-series eigenvalue of the transaction order, represents the logistics node lag time value, represents the abnormal trading behavior volume, represents the total number of high-difference feature intervals;

[0020] S104: Perform sparsification processing on the abnormal trading behavior impact value set, mark the intervals where the abnormal trading behavior impact value is greater than the sparsity threshold as valid intervals, and generate a sparse feature vector set by extracting the feature parameters in the valid intervals.

[0021] As a further solution of the present invention, the obtaining step of the global key node trend prediction value is specifically as follows:

[0022] S201: Extract and calculate the normalized values of multiple time series in the sparse feature vector from the sparse feature vector set, calculate the dynamic change rate of time points for the normalized value sequence, filter out the prominent change time points based on the dynamic change rate fluctuation value, and generate a global trade data change trend sequence value;

[0023] S202: Compare the global trade data change trend sequence value with the dynamic trend of the core nodes of the trading network, extract the nodes with significant sequence differences, analyze the dynamic trend of the nodes with significant differences in segments, and calculate the mean square error value based on the difference between the node rate and the global dynamic trend rate, and generate the dynamic trend of the core trading nodes;

[0024] S203: Match the lag time of the key nodes in the dynamic trend of the core transaction nodes with the standard, calculate the deviation value of the lag time, using the formula:

[0025] ;

[0026] Extract the nodes with larger lag time deviation values as the nodes with lag exceeding the standard, and generate the key nodes with lag time exceeding the standard;

[0027] Among them, represents the deviation value of the lag time of the key nodes, represents the lag time of the dynamic trend nodes, represents the standard time series, represents the total number of nodes, is the adjustment weight, for the scaling adjustment effect of the lag time standard deviation, represents the standard deviation of the lag time, measuring the degree of dispersion of the time series, represents the summation operation for all nodes;

[0028] S204: Analyze the change of the connectivity weight between the key nodes with lag time exceeding the standard and the core transaction nodes, calculate the connectivity weight fluctuation value based on the weight change rate of each node, screen the key nodes whose fluctuation value is related to the lag impact, and establish the global key node trend prediction value.

[0029] As a further solution of the present invention, the steps for obtaining the distributed key node dynamic weight matrix are specifically as follows:

[0030] S301: Based on the global key node trend prediction value, refine the time change data of each node, extract the time fine-grained changes of multiple nodes through time series decomposition, analyze the change trend in combination with the time point information, and generate the time fine-grained node change data;

[0031] S302: Perform time series correlation measurement on the time fine-grained node change data, calculate the time correlation coefficient between adjacent nodes based on the time series of the nodes, screen the nodes with strong correlation through correlation analysis, and generate the time series correlation measurement value of each node;

[0032] S303: Extract the time change rate in the time series correlation measurement value, calculate the change rate of the node time series through the difference algorithm, screen and mark the nodes with large rate differences, and generate the time change rate value;

[0033] S304: Compare the time change rate value with the connectivity weight value of adjacent nodes, using the formula:

[0034] ;

[0035] Call the node update rule, adjust the weights node by node, and establish a distributed dynamic weight matrix for key nodes;

[0036] Among them, represents the updated node weight, represents the original node weight, is the adjustment coefficient, used to control the impact of rate changes on weight adjustment, represents the rate difference between nodes, is the average value of the node rates.

[0037] As a further solution of the present invention, the steps for obtaining the node correlation risk transmission index are specifically as follows:

[0038] S401: Extract the risk factor combination value between nodes based on the distributed dynamic weight matrix for key nodes. By analyzing the dynamic change rate of multiple nodes in the weight matrix and combining the risk factor influence value, calculate the risk correlation between multiple nodes, and generate the risk factor combination value between nodes;

[0039] S402: Calculate the conduction probability of the risk factor combination value between nodes. Select node groups through risk correlation, analyze the transmission path and probability distribution of risk factors between nodes, and combine the cumulative transmission model to calculate the risk conduction probability, and generate the risk conduction probability value between nodes;

[0040] S403: Compare and analyze the risk conduction probability value between nodes with the abnormal change area of the weight matrix, calculate the correlation between the risk conduction probability and the weight change, and use the formula:

[0041] ;

[0042] Weight the risk conduction probability in combination with the weight change rate, and screen out the nodes with significant risks;

[0043] Among them, represents the correlation degree value of the nodes with significant risks, is the conduction probability of the th node, is the weight value of the th node, is the node weight change amount, is the average node weight, is the adjustment factor to control the impact of weight change;

[0044] S404: Analyze the dynamic weight difference of the nodes with significant risks. By comparing the matching degree between the dynamic weight changes between nodes and the conduction probability value, calculate the risk transmission relationship between nodes in combination with the risk correlation degree, and establish the node correlation risk transmission index.

[0045] As a further solution of the present invention, the steps for obtaining the set of optimized paths for key nodes are specifically as follows:

[0046] S501: Based on the node - associated risk transmission index, by analyzing the risk conduction probability and weight change value between nodes node - by - node, screening the key nodes with higher conduction probability among the risk nodes, determining the initial optimized path in combination with the path flow characteristics, and generating an optimized path sequence for the risk nodes;

[0047] S502: Adjust the critical path flow value and weight value of the optimized path sequence of the risk nodes. Calculate the contribution weights of multiple nodes based on the proportion of node flow in the path, re - distribute the flow value through the total weight difference rate, calculate the adjusted path connectivity index, and adjust the critical path flow value and weight value;

[0048] S503: Verify the connectivity of the optimized path after adjusting the critical path flow value and weight value. By calculating the impact of the flow distribution ratio on the path connectivity, using the formula:

[0049] ;

[0050] Optimize the standardized distribution of node flows on the path, generate an optimized flow distribution in combination with the weight adjustment ratio, and adjust the flow distribution ratio;

[0051] Wherein, represents the path connectivity index, measuring the influence of the flow distribution and weight among nodes in the path, is the flow value of the th node on the path, representing the transmission capacity of the node, is the weight value corresponding to the node, representing the importance of the node, is the total number of nodes on the path, providing a reference for the path scale, is the adjustment factor, used to control the influence of flow fluctuations on the path connectivity, represents the standard deviation of the path node flows, measuring the uniformity of the node flow distribution;

[0052] S504: Analyze the optimization characteristics of the path after adjusting the flow distribution ratio. By calculating the flow balance degree and node connectivity matching rate within the path, screening the paths according to the balanced flow distribution and connectivity, and generating a set of optimized paths for key nodes according to the optimization rules.

[0053] A digital optimization system for trade processes, the digital optimization system for trade processes is used to execute the above - mentioned digital optimization method for trade processes, and the system includes:

[0054] Based on the trade data set, the data feature extraction module extracts the time series feature values of transaction orders, the lag time values of logistics nodes, and the abnormal transaction behavior volume, calculates the differences of multi-feature values, calculates the range of transaction behavior features, compares and adjusts the feature screening rules, optimizes the correlation features between transaction behaviors and logistics nodes, and generates a sparse feature vector set;

[0055] Based on the sparse feature vector set, the global trend analysis module calculates the global trade data change trend sequence value, extracts the dynamic change trend of core transaction nodes, determines whether the lag time value of key nodes meets the standard, analyzes the change of connectivity weights of core transaction nodes, and verifies the consistency to establish the global key node trend prediction value;

[0056] Based on the global key node trend prediction value, the key node optimization module separates the time change data of fine-grained nodes, calculates the time series correlation metric value of each node, extracts the time change rate value, compares the connectivity weight values of adjacent nodes, calls the node update rule to iteratively optimize the node correlation, and establishes a distributed key node dynamic weight matrix;

[0057] Based on the distributed key node dynamic weight matrix, the risk factor analysis module extracts the risk factor combination values between nodes, calculates the risk conduction probability value, analyzes the abnormal change area in the weight matrix, screens the nodes with significant risks, compares the dynamic weight differences between nodes, and establishes a node correlation risk transmission index;

[0058] Based on the node correlation risk transmission index, the path optimization module constructs an optimized path sequence of risk nodes, adjusts the flow values and weight values of key paths, verifies the connectivity of the optimized paths, adjusts the flow distribution ratio, and generates a set of optimized paths for key nodes.

[0059] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0060] In the present invention, by deeply analyzing transaction data, extracting the time series features of transaction orders and the lag time of logistics nodes, accurately screening transaction behavior features, optimizing data correlation, and improving the coordination of the transaction chain. Based on the global data trend analysis, dynamically identifying the changes of key nodes, ensuring time series coherence and optimizing the connectivity between nodes. Through fine-grained time series analysis, enhancing the cooperation ability between nodes, reducing the impact of abnormal delays. Constructing a dynamic weight matrix, real-time monitoring and identifying potential risks, adjusting the path flow and weight distribution, realizing the optimized allocation of resources, reducing the supply chain pressure, and improving the overall operation efficiency. Brief Description of the Drawings

[0061] Figure 1 is a schematic diagram of the working process of the present invention;

[0062] Figure 2Flow chart of the acquisition steps of the sparse feature vector set of the present invention;

[0063] Figure 3 Flow chart of the acquisition steps of the global key node trend prediction value of the present invention;

[0064] Figure 4 Flow chart of the acquisition steps of the distributed key node dynamic weight matrix of the present invention;

[0065] Figure 5 Flow chart of the acquisition steps of the node association risk transmission index of the present invention;

[0066] Figure 6 Flow chart of the acquisition steps of the key node optimization path set of the present invention. Detailed implementation manners

[0067] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0068] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, in the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined. Embodiment 1

[0069] Please refer to Figure 1 , the present invention provides a technical solution: a method for digital optimization of trade processes, including the following steps:

[0070] S1: Extract the time series feature values, logistics node lag time values, and abnormal transaction behavior amounts of transaction orders from the trade data set, extract the transaction behavior feature range through feature difference operations, compare the difference values between the data, adjust the feature screening rules to optimize the association features of transaction behaviors and logistics nodes, and generate a sparse feature vector set;

[0071] S2: Based on the sparse feature vector set, calculate the global trade data change trend sequence value, extract the dynamic trend of the core transaction nodes, judge whether the lag time of the key nodes meets the standard, analyze the change of the connectivity weight of the core transaction nodes, verify the consistency, and establish a global key node trend prediction value;

[0072] S3: Based on the global key node trend prediction value, separate the fine-grained node time-varying data, calculate the temporal correlation metric value of each node, extract the time change rate value, compare the connectivity weight values of adjacent nodes, call the node update rule to perform association iterative optimization, and establish a distributed key node dynamic weight matrix;

[0073] S4: Based on the distributed key node dynamic weight matrix, extract the risk factor combination value between nodes, calculate the risk conduction probability value, analyze the abnormal change area of the weight matrix, screen the risk-significant nodes, compare the dynamic weight differences between nodes, and establish a node association risk transfer index;

[0074] S5: Based on the node association risk transfer index, construct an optimized path sequence for risk nodes, adjust the key path traffic value and weight value, verify the connectivity of the optimized path, adjust the traffic distribution ratio, and generate a set of optimized paths for key nodes.

[0075] The sparse feature vector set includes the trading behavior feature range, the optimized value of the feature screening rule, and the association features between trading behavior and logistics nodes. The global key node trend prediction value includes the dynamic change trend of core trading nodes, the standard value of key node lag time, and the change of connectivity weight of core trading nodes. The distributed key node dynamic weight matrix includes the fine-grained node time-varying data, the temporal correlation metric value, the time change rate value, and the connectivity weight value of adjacent nodes. The node association risk transfer index includes the risk factor combination value between nodes, the risk conduction probability value, the abnormal change area of the weight matrix, the risk-significant nodes, and the dynamic weight difference between nodes. The set of optimized paths for key nodes includes the optimized path sequence for risk nodes, the key path traffic value, the weight value, and the traffic distribution ratio.

[0076] Please refer to Figure 2 , and the specific steps for obtaining the sparse feature vector set are as follows:

[0077] S101: Extract the temporal feature values of trading orders from the trade data set, sort the extracted temporal feature values, generate a time difference sequence by calculating the time difference between trading orders, screen the trading orders in the high-frequency trading time period according to the distribution characteristics of the time difference sequence, calculate the intersection ratio between the high-frequency trading time period and the overall trading time series, and generate a trading high-frequency feature interval;

[0078] Extract the time series feature values of transaction orders from the trade data set. First, classify the order data in chronological order, generate a time difference sequence by calculating the time difference between each transaction order, and use the time difference sequence as the basic feature quantity to measure the trading time interval. Then, analyze the distribution characteristics of the time difference sequence, and use the cumulative distribution function (CDF) to map the time difference sequence to different time period intervals, so as to determine the distribution range of high-frequency trading time periods. After that, calculate the intersection ratio of the number of transaction orders in the high-frequency trading time period and the total number of transaction orders. The calculation method is the ratio formula: , where is the number of orders in the high-frequency time period, is the total number of transaction orders. Finally, combine the calculated intersection ratio result with the time period feature value to obtain and mark the range feature of the high-frequency trading time period, and generate the high-frequency trading feature interval.

[0079] S102: Cross-compare the time period feature values in the high-frequency trading feature interval with the logistics node lag time value, calculate the average difference and standard deviation of each time period by interval, and screen out the intervals where the average difference is greater than a certain multiple of the standard deviation. Generate a set of high-difference feature intervals by comparing the interval coverage rate;

[0080] Compare the time period feature values in the high-frequency trading feature interval with the logistics node lag time value. By calculating the absolute difference between the time feature value of the high-frequency trading time period and the logistics node lag time one by one, calculate the average value and standard deviation of the absolute difference calculation results of each time period. The formulas are the average value formula: and the standard deviation formula: . Screen out the time periods where the average absolute difference is greater than a certain multiple of the standard deviation to form high-difference feature intervals. Generate a set of high-difference feature intervals by superimposing and comparing the coverage ranges of the high-difference intervals with the comparison features of subsequent transaction node information.

[0081] S103: Cumulatively calculate the abnormal trading behavior volume in the set of high-difference feature intervals, using the formula:

[0082] ;

[0083] Calculate the influence value of abnormal trading behavior in multiple intervals to generate a set of influence values of abnormal trading behavior;

[0084] Among them, represents the influence value of abnormal trading behavior, represents the time series feature value of the transaction order, represents the logistics node lag time value, represents the abnormal trading behavior volume, represents the total number of high-difference feature intervals;

[0085] Formula:

[0086] ;

[0087] The benefit of the formula is that by combining the time series characteristic values of transaction orders, the lag time values of logistics nodes, and the volume of abnormal transaction behaviors, it can accurately quantify the impact values of abnormal transaction behaviors in each time period, providing precise data support for subsequent optimization.

[0088] Detailed explanation of the formula and the derivation process of formula calculation:

[0089] represents the time series characteristic value of the transaction order, obtained by monitoring the daily transaction time points; represents the lag time value of the logistics node, calculated by extracting the difference between the actual processing time and the planned time of the corresponding node; is the volume of abnormal transaction behaviors, which needs to screen and accumulate the transaction behavior data and is obtained by calculating the total number of transaction behaviors that do not meet the expectations every day; is the number of high-difference feature intervals, obtained by counting the high-difference feature interval set generated in step 2.

[0090] Assignment: Assume is [10, 12, 14] hours, is [8, 11, 13] hours, is [3, 5, 4], , formula derivation:

[0091] ;

[0092] The result shows that the impact value of abnormal transaction behaviors in the time period is 5. By analyzing the cumulative impact of each transaction interval, it can provide a reference for the next sparsification process.

[0093] S104: Sparsify the set of impact values of abnormal transaction behaviors, mark the intervals where the impact values of abnormal transaction behaviors are greater than the sparsity threshold as valid intervals, and generate a set of sparse feature vectors by extracting the feature parameters in the valid intervals.

[0094] Sparsify the set of impact values of abnormal transaction behaviors. Screen the impact values of abnormal transaction behaviors according to a certain sparsity threshold. First, calculate the distribution density of the set of impact values of abnormal transaction behaviors, divide the value range interval through the distribution density function, extract the intervals where the impact values are greater than a specific sparsity threshold as valid intervals, then extract and analyze the transaction order feature parameters in each valid interval, and finally construct a set of sparse feature vectors by minimizing the parameter dimension.

[0095] Please refer to Figure 3 , the steps for obtaining the predicted value of the global key node trend are specifically as follows:

[0096] S201: Extract and calculate the normalized values of multiple time series in the sparse feature vector from the sparse feature vector set, calculate the dynamic change rate of time points for the normalized value sequence, screen out the prominent change time points based on the fluctuation value of the dynamic change rate, and generate the global trade data change trend sequence value;

[0097] Extract the global trade data change trend sequence value from the sparse feature vector set, calculate the normalized values of each time series in the sparse feature vector. The normalized value is obtained by performing point-by-point normalization on the features of the time point sequence to ensure that the features are within the same dimension for change rate calculation. For the dynamic change rate calculation, the value difference between consecutive time points in the time series is used as the numerator, and the normalized time point value is used as the denominator to obtain the change rate value. To screen the change rate fluctuation value, it is necessary to classify according to the difference amount and set a fluctuation threshold based on the normalized range. The significant fluctuation points are screened by marking and extracting the points with a fluctuation value greater than the threshold. After completing the above steps, the marked significant fluctuation points are summarized into a set of significant change time points to generate the global trade data change trend sequence value.

[0098] S202: Compare the global trade data change trend sequence value with the dynamic trend of the core nodes of the trading network, extract the nodes with significant sequence differences, analyze the dynamic trend of the nodes with significant differences in segments, calculate the mean square error value based on the difference between the node rate and the global dynamic trend rate, and generate the dynamic trend of the core trading nodes;

[0099] Compare the global trade data change trend sequence value with the dynamic trend of the core nodes of the trading network. First, the dynamic trend of the core nodes needs to be realized by extracting the time series features of the key nodes in the trading network. The extraction of the dynamic trend requires segmenting the time series and calculating the change rate of each segment. The rate value is obtained by dividing the difference between the feature values of the two end time points by the time span of the time period. When extracting the nodes with significant sequence differences, it is necessary to calculate the mean square error value between the change rate of each node and the global dynamic trend rate. The screening of significant nodes is achieved by setting the difference threshold of the mean square error. The segment analysis requires refining the segment processing of the time series of significant nodes and analyzing in combination with the rate change characteristics of each segment to generate the dynamic trend of the core trading nodes.

[0100] S203: Perform standard matching on the lag time of the key nodes in the dynamic trend of the core trading nodes, calculate the deviation value of the lag time, using the formula:

[0101] ;

[0102] Extract the nodes with relatively large lag time deviation values as the nodes with lag exceeding the standard, and generate the key nodes with lag time exceeding the standard;

[0103] Among them, represents the deviation value of the lag time of the key node, represents the lag time of the dynamic trend node, represents the standard time series, represents the total number of nodes, is the adjustment weight for the scaling adjustment effect of the lag time standard deviation, represents the standard deviation of the lag time, measuring the degree of dispersion of the time series, represents the summation operation for all nodes;

[0104] Formula:

[0105] ;

[0106] The benefit of the formula is that by adding the joint action of the adjustment weight and the standard deviation , the sensitivity of the deviation value to the distribution characteristics and outliers is improved, so as to more accurately evaluate the deviation of the lag time of the key node.

[0107] Detailed explanation of the formula and the derivation process of the formula calculation:

[0108] Let the lag time be , the standard time be , the weight , calculate the standard deviation , the summation result be , and the deviation value is calculated as:

[0109] ;

[0110] This result shows that the deviation value of the lag time is 0.37, reflecting the degree of lag time deviation of the key node, which can be further used to screen the key nodes with relatively large lag time deviation as the analysis object and generate the key nodes with lag time exceeding the standard.

[0111] S204: Analyze the change of the connectivity weight between the key nodes with lag time exceeding the standard and the core transaction nodes, calculate the connectivity weight fluctuation value based on the weight change rate of each node, screen the key nodes whose fluctuation value is associated with the lag effect, and establish the global key node trend prediction value.

[0112] Conduct a comprehensive analysis of the connectivity weight changes of key nodes with excessive lag time and core transaction nodes. The nodes with excessive lag time calculate the correlation through their lag time values and weight change rates. To obtain the connectivity weight changes, it is necessary to record the trading volume characteristics of each node one by one and normalize the trading volume change sequence. The normalized change sequence is used to calculate the weight change rate of the node. The rate value is calculated by dividing the change amount between the two points before and after the sequence by the time point interval. The weight change fluctuation value is calculated according to the correlation fitting function of the rate and the lag time. When screening nodes with significant fluctuation values, a screening threshold is set based on the fitting residual value, and a global key node trend prediction value is established.

[0113] Please refer to Figure 4 , and the steps to obtain the dynamic weight matrix of distributed key nodes are specifically as follows:

[0114] S301: Based on the global key node trend prediction value, refine the time change data of each node, extract the time fine-grained changes of multiple nodes through time series decomposition, analyze the change trend in combination with the time point information, and generate time fine-grained node change data;

[0115] Based on the global key node trend prediction value, extract the time change data of each node. By constructing the time series of the node, decompose the time series data into fine-grained time change points according to the node. Combine the attribute data of the node to further subdivide the eigenvalue of each time point in the time series. For the change rate between different time points, compare the growth or decay trend of the time series between nodes, classify the time series by the change amplitude value, generate the change trend curve of the fine-grained time point through segmented accumulation, combine the time point correlation relationship between nodes, integrate the data according to the time segment, obtain the change data of the node time segment, filter out the node time data with fluctuation characteristics based on the data, and extract the unique trend of the key node in the time change to generate time fine-grained node change data;

[0116] S302: Conduct a time series correlation measurement on the time fine-grained node change data, calculate the time correlation coefficient between adjacent nodes based on the time series of the node, screen out the nodes with strong correlation through correlation analysis, and generate the time series correlation measurement value of each node;

[0117] Perform temporal correlation measurement on the time fine-grained node change data. Based on the time point distribution between adjacent nodes in the time series, construct a correlation matrix. By calculating the correlation coefficient of the time series between nodes, screen the node pairs with significant correlation, mark the significantly correlated node relationships in the correlation matrix as strongly correlated, calculate the difference degree value of the time series by combining the time point trend curve of the nodes, judge the temporal correlation between nodes by setting a threshold for the difference degree value, further establish a time correlation model based on the time distribution characteristics of the nodes, adjust the node connection relationship in the model, update the time correlation coefficient of the nodes according to the distribution of the correlation degree matrix, and finally integrate the time point data and correlation coefficient of each node to generate the temporal correlation measurement value for each node;

[0118] S303: Extract the time change rate in the temporal correlation measurement value, calculate the change rate of the node time series through the difference algorithm, screen and mark the nodes with large rate differences, and generate the time change rate value;

[0119] Extract the time change rate in the temporal correlation measurement value, calculate the change rate of the time series of each node through the difference algorithm. In view of the volatility of the node time change rate, compare the rate differences of each node, combine the rate fluctuation value to screen out the nodes with large rate differences, set a rate change threshold, mark the nodes with rate changes exceeding the threshold as rate abnormal nodes, classify the time series fluctuation characteristics of the rate abnormal nodes, generate a set of nodes with significant rate differences through classification statistics, calculate the slope value of the rate change curve, analyze the change trend of the time rate of each node based on the slope value, and integrate the rate change trends of the abnormal nodes to generate the time change rate value;

[0120] S304: Compare the time change rate value with the connection weight value of adjacent nodes, and use the formula:

[0121] ;

[0122] Call the node update rule, adjust the weight node by node, and establish a distributed key node dynamic weight matrix;

[0123] Among them, represents the updated node weight, represents the original node weight, is the adjustment coefficient, which is used to control the impact of rate changes on weight adjustment, represents the rate difference between nodes, is the average value of the node rates.

[0124] Formula:

[0125] ;

[0126] The advantage of the formula is that by introducing the ratio of the rate change to the average rate and combining it with the adjustment coefficient , it effectively reflects the dynamic impact of the rate change between adjacent nodes on weight adjustment, and improves the sensitivity and accuracy of weight update;

[0127] Detailed explanation of the formula and the derivation process of formula calculation:

[0128] Assume that the initial node weight is , the adjustment coefficient , the node rate change , the average node rate , substitute into the formula:

[0129] ;

[0130] Calculate the internal ratio: ;

[0131] Calculate the adjustment amount: ;

[0132] Calculate the weight update value: ;

[0133] This result shows that the node weight is updated from the initial value to . The updated weight value comprehensively considers the rate change and node dynamic characteristics, which helps to construct a distributed key node dynamic weight matrix.

[0134] Please refer to Figure 5 . The specific steps for obtaining the node association risk transmission index are as follows:

[0135] S401: Extract the risk factor combination value between nodes based on the distributed key node dynamic weight matrix. By analyzing the dynamic change rate of multiple nodes in the weight matrix and combining the risk factor influence value, calculate the risk correlation between multiple nodes, and generate the risk factor combination value between nodes;

[0136] Extract the risk factor combination value between nodes based on the distributed key node dynamic weight matrix. Analyze the dynamic change rate of each node in the weight matrix. By calculating the relative amplitude and cumulative value of each change in the time series for each node one by one, use the formula: Quantify the change situation of each node. represents the weight change value of the th node in the weight matrix. is the average value of all node weight values. According to the calculation result of the weight change rate, the node range with a change rate higher than the threshold is screened out; combined with the influence range data of risk factors, the risk factor combinations are matched item by item for the high-change-rate nodes, and the risk factors are quantified into influence probability values. For example, the economic fluctuation factor ranges from 0.3 to 0.5, and the market delay factor ranges from 0.1 to 0.2. According to the dynamic change ratio, the risk correlation is matched, and the risk correlation combination value between two nodes is calculated. The average value of the dynamic change rate is multiplied by the average value of the risk factor influence range, and the combined correlation weight is calculated by adjusting the multiple. The weight value is stored as the quantified result of the risk factor combination, and the risk factor combination value between nodes is generated.

[0137] represents the weight change rate of the

[0138] th node, which is used to quantify the relative degree of weight change of the node in the time series;

[0139] represents the average value of all node weights, which is used to standardize the influence of weight change. The calculation formula is: ;

[0140] where is the weight value of the th node,

[0141] and

[0142] is the total number of nodes; If the node weights are

[0143] ;

[0144] Substitute into the formula: ;

[0145] The calculation result is: ;

[0146] Match the combined weight according to the risk factor influence range. For example, if the economic fluctuation factor is 0.4 and the market delay factor is 0.15, calculate the combined correlation weight for the high-change-rate nodes:​​​​​ ;

[0147] Calculated: ;

[0148] The result shows that the correlation between the node change rate and the risk factor is quantified as 0.09005, which can be further used for the analysis of risk transmission between nodes.

[0149] S402: Calculate the conduction probability of the combined values of risk factors between nodes, select node groups through risk correlation, analyze the transmission path and probability distribution of risk factors between nodes, and calculate the risk conduction probability by combining the cumulative transmission model to generate the risk conduction probability value between nodes;

[0150] Calculate the conduction probability of the combined values of risk factors between nodes, identify node groups with relatively high risk correlation, analyze the conduction path and probability distribution pair by pair. When calculating the conduction probability value of each node, quantify the combined value of risk factors as the input parameter, and calculate through the formula where is the influence weight of the risk factor, is the node risk change rate. By monitoring the total risk of each node in the conduction path, screen the key nodes in the conduction path, and judge the conduction probability size of each node according to the distribution interval of the conduction probability value. For example, the conduction probability calculation of multiple nodes gives , , . Mark the nodes with values lower than the average as nodes with weak conduction, extract the path combinations of nodes with high conduction probability, calculate the comprehensive conduction value of the key nodes using the total conduction path, and generate the risk conduction probability value between nodes.

[0151] S403: Compare and analyze the risk conduction probability value between nodes with the abnormal change region of the weight matrix, calculate the correlation between the risk conduction probability and the weight change, and use the formula:

[0152] ;

[0153] Weight the risk conduction probability in combination with the weight change rate to screen out the nodes with significant risks;

[0154] where represents the correlation degree value of the nodes with significant risks, is the conduction probability of the th node, is the weight value of the th node, is the node weight change amount, is the average node weight, is the adjustment factor to control the influence of the weight change;

[0155] Formula:

[0156] ;

[0157] The benefit of the formula is that by the combined effect of dynamically adjusting the conduction probability, the weight change rate, and the adjustment factor, it can accurately capture the change characteristics of risk-significant nodes and improve the system's precise quantification ability of risk conduction.

[0158] Detailed explanation of the formula and the derivation process of formula calculation:

[0159] Obtained by monitoring the risk conduction path, Collected through the weight matrix, Calculated through the weight change value, Is the average value of the node weights, Is the adjustment factor, substitute it into the formula:

[0160] ;

[0161] Calculated item by item:

[0162] ;

[0163] ;

[0164] This result shows that the comprehensive correlation degree of risk-significant nodes is 1.270, which exceeds the range compared with the benchmark value of 0.8, indicating that these nodes have higher risks in weight changes and conduction paths, and are further used to screen risk-significant nodes.

[0165] S404: Analyze the dynamic weight differences of risk-significant nodes, calculate the risk transfer relationship between nodes by comparing the matching degree between the dynamic weight changes and the conduction probability values among nodes, and establish a node-associated risk transfer index.

[0166] Analyze the dynamic weight differences of risk-significant nodes. By pairwise comparing the dynamic weight changes of nodes and combining the conduction probability values to evaluate the matching relationship between nodes, for nodes with significant dynamic weight changes, use the node with a higher change rate as the initial input parameter, calculate the cumulative value of the conduction path between nodes, synthesize the weight differences in the conduction path, calculate the risk transfer correlation degree of each node through the cumulative value of the dynamic conduction path, map the correlation degree of the node to a specific transfer index value, pairwise adjust the transfer order of nodes to optimize the final index, and establish a node-associated risk transfer index.

[0167] Please refer to Figure 6 , and the specific steps for obtaining the key node optimization path set are as follows:

[0168] S501: Based on the node - associated risk transfer index, by analyzing the risk conduction probability and weight change value between nodes node - by - node, screening the key nodes with higher conduction probability among the risk nodes, and determining the initial optimized path in combination with the path flow characteristics, generating an optimized path sequence for the risk nodes;

[0169] Based on the node - associated risk transfer index, by analyzing the risk conduction probability and weight change value between nodes node - by - node, first collect the conduction probability values of each node on the path, use the monitoring system to record the data flow values transmitted between nodes, sort the risk conduction probabilities in combination with the dynamic weight change rate, and select the top 20% of the nodes as the key node group. Then, analyze the flow fluctuation amplitude and node connectivity in the key node group, screen the nodes with poor flow stability through the standard deviation calculation formula of the fluctuation amplitude, preferentially include the nodes with poor stability into the initial optimized path. Finally, adjust the flow proportion between nodes according to the balance criterion of path flow, and determine the initial optimized path according to the standardized results of all node flows on the path, generating an optimized path sequence for the risk nodes.

[0170] S502: Adjust the critical path flow value and weight value of the optimized path sequence of the risk nodes, calculate the contribution weights of multiple nodes based on the node flow proportion within the path, redistribute the flow values through the total weight difference rate, calculate the adjusted path connectivity index, and adjust the critical path flow value and weight value;

[0171] Adjust the critical path flow value and weight value of the optimized path sequence of the risk nodes. Calculate the contribution weights of each node based on the node flow proportion within the path. First, count the node flow values in the optimized path, use the standardization processing formula to calculate the relative flow proportion values of each node within the path, multiply the proportion values by the initial weight values of the nodes to calculate their contribution weights. Then, compare the difference rate between the total weight of the nodes within the path and the target weight, redistribute the flow values of the nodes within the path according to the weight difference ratio. At the same time, use the connectivity calculation formula to verify the overall connectivity level of the path after adjustment, and compare the adjusted flow values with the initial values to confirm the rationality of the distribution, and adjust the critical path flow value and weight value.

[0172] S503: Verify the connectivity of the optimized path after adjusting the critical path flow value and weight value, by calculating the impact of the flow distribution ratio on the path connectivity, using the formula:

[0173] ;

[0174] Optimize the standardized distribution of the node flows on the path, generate an optimized flow distribution in combination with the weight adjustment ratio, and adjust the flow distribution ratio;

[0175] Wherein, Represents the path connectivity index, which measures the influence of the traffic distribution and weights among nodes within the path. Is the traffic value of the -th node on the path, indicating the transmission capacity of this node. Is the weight value corresponding to the node, indicating the importance of the node. Is the total number of nodes on the path, providing a reference for the path scale. Is the adjustment factor, used to control the influence of traffic fluctuations on the path connectivity. Represents the standard deviation of the traffic of path nodes, measuring the uniformity of the node traffic distribution;

[0176] Formula:

[0177] ;

[0178] The benefit of the formula is that by adding the standard deviation term of node traffic to reflect the traffic distribution uniformity, introducing the adjustment factor to dynamically adjust the weight of traffic fluctuations on the connectivity index, enhancing the sensitivity of path connectivity optimization.

[0179] Detailed explanation of the formula and the derivation process of formula calculation:

[0180] ;

[0181] Among them, assume that the traffic values of 5 nodes on the path are respectively , , , , , and the node weight values are respectively , , , , , the traffic standard deviation is calculated to be 25, and the adjustment factor , the calculation process is as follows:

[0182] 1. Calculate the sum of the products of weights and traffic: ;

[0183] 2. Calculate the sum of traffic: ;

[0184] 3. Calculate the adjustment term: ;

[0185] 4. Calculate the connectivity index: ;

[0186] The result shows that the path connectivity index is 0.791, close to 1, indicating that both the path flow and weight distribution are relatively balanced. The balance of the flow distribution can be further improved through optimization, and finally, the stable connectivity of the path can be optimized.

[0187] S504: Conduct an optimization feature analysis on the path after adjusting the flow distribution ratio. By calculating the flow balance degree within the path and the node connectivity matching rate, screen the paths according to the balanced flow distribution and connectivity, and generate a set of optimized paths for key nodes according to the optimization rules.

[0188] Conduct an optimization feature analysis on the path after adjusting the flow distribution ratio. By calculating the flow balance degree within the path and the node connectivity matching rate, first, for the adjusted node flow distribution, calculate the balance using the flow variance formula. A smaller variance indicates a more balanced distribution. Select the paths with the flow variance controlled within 5% as candidates. Subsequently, calculate the connectivity matching rate between nodes. A matching rate greater than 90% indicates a higher path stability. Screen the paths with a higher matching rate to generate an optimized path set. Finally, conduct a cross-analysis on the paths that meet the conditions for both balance and connectivity, and select the intersection part as the final result to generate a set of optimized paths for key nodes.

[0189] A digital optimization system for trade processes, which is used to execute the above digital optimization method for trade processes. The system includes:

[0190] The data feature extraction module extracts the time series feature values of transaction orders, the lag time values of logistics nodes, and the abnormal transaction behavior volume based on the trade data set, calculates the differences of multiple feature values, calculates the range of transaction behavior features, compares and adjusts the feature screening rules, optimizes the association features between transaction behaviors and logistics nodes, and generates a set of sparse feature vectors;

[0191] The global trend analysis module calculates the global trade data change trend sequence values based on the set of sparse feature vectors, extracts the dynamic change trends of core transaction nodes, judges whether the lag time values of key nodes meet the standards, analyzes the changes in the connectivity weights of core transaction nodes, and verifies the consistency to establish the global key node trend prediction values;

[0192] The key node optimization module separates the fine-grained node time change data based on the global key node trend prediction values, calculates the time series correlation metric values of each node, extracts the time change rate values, compares the connectivity weight values of adjacent nodes, and calls the node update rules to iteratively optimize the node association, and establishes a distributed key node dynamic weight matrix;

[0193] The risk factor analysis module extracts the combined risk factor values between nodes based on the distributed critical node dynamic weight matrix, calculates the risk conduction probability values, analyzes the abnormal change regions in the weight matrix, screens the nodes with significant risks, compares the dynamic weight differences between nodes, and establishes the node correlation risk transfer index;

[0194] The path optimization module constructs the optimized path sequence of the risk nodes based on the node correlation risk transfer index, adjusts the flow values and weight values of the critical paths, verifies the connectivity of the optimized paths, adjusts the flow distribution ratio, and generates the set of optimized paths for the critical nodes.

[0195] The above is only the preferred embodiment of the present invention, and it is not intended to limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for digital optimization of trade processes, characterized in that, It includes the following steps: S101: Extract the time series feature values of the transaction orders from the trade data set, sort the extracted time series feature values, generate a time difference sequence by calculating the time differences between the transaction orders, filter the transaction orders in the high-frequency trading time period according to the distribution characteristics of the time difference sequence, calculate the intersection ratio of the high-frequency trading time period and the overall trading time series, and generate a trading high-frequency feature interval; S102: Cross-compare the time period feature values in the trading high-frequency feature interval with the logistics node lag time values, calculate the difference mean and difference standard deviation for each period by interval, filter the intervals where the difference mean is greater than a certain multiple of the standard deviation, and generate a high-difference feature interval set by comparing the interval coverage rate; S103: Cumulatively calculate the abnormal trading behavior amounts in the high-difference feature interval set, and use the formula: ； Calculate the abnormal trading behavior influence values for multiple intervals, and generate a set of abnormal trading behavior influence values; Among them, represents the impact value of abnormal trading behavior, represents the time series characteristic value of the trading order, represents the lag time value of the logistics node, represents the volume of abnormal trading behavior, represents the total number of high-difference characteristic intervals; S104: Perform sparsification processing on the set of abnormal trading behavior influence values, mark the intervals where the abnormal trading behavior influence values are greater than the sparsity threshold as valid intervals, and generate a set of sparse feature vectors by extracting the feature parameters in the valid intervals; S201: Extract and calculate the normalized values of multiple time series in the sparse feature vectors from the set of sparse feature vectors, calculate the dynamic change rate of time points for the normalized value sequence, filter the prominent change time points based on the fluctuation value of the dynamic change rate, and generate the global trade data change trend sequence values; S202: Compare the global trade data change trend sequence values with the dynamic trend of the core nodes of the trading network, extract the nodes with significant sequence differences, analyze the dynamic trend of the nodes with significant differences in segments, calculate the mean square error value based on the difference between the node rate and the global dynamic trend rate, and generate the dynamic trend of the core trading nodes; S203: Perform standard matching on the lag time of the key nodes in the dynamic trend of the core trading nodes, calculate the deviation value of the lag time, and use the formula: ； Extract the nodes with larger lag time deviation values as the nodes with excessive lag, and generate the key nodes with excessive lag time; Among them, represents the deviation value of the lag time of the key node, represents the lag time of the dynamic trend node, represents the standard time series, represents the total number of nodes, is the adjustment weight for the scaling adjustment effect of the standard deviation of the lag time, represents the standard deviation of the lag time, measuring the degree of dispersion of the time series, represents the summation operation for all nodes; S204: Analyze the change in the connectivity weight between the key nodes with excessive lag time and the core trading nodes, calculate the connectivity weight fluctuation value based on the weight change rate of each node, filter the key nodes where the fluctuation value is associated with the lag effect, and establish the global key node trend prediction value; S3: Based on the global key node trend prediction value, separate the fine-grained node time change data, calculate the time series correlation metric value for each node, extract the time change rate value, compare the connectivity weight values of adjacent nodes, call the node update rule to perform association iterative optimization, and establish a distributed key node dynamic weight matrix; S4: Based on the distributed key node dynamic weight matrix, extract the combined values of the risk factors between nodes, calculate the risk conduction probability value, analyze the abnormal change area of the weight matrix, filter the nodes with significant risks, compare the dynamic weight differences between nodes, and establish a node association risk transmission index; S5: Based on the node - associated risk transfer metrics, construct an optimized path sequence for risk nodes, adjust the traffic values and weight values of critical paths, verify the connectivity of the optimized paths, adjust the traffic distribution ratio, and generate a set of optimized paths for critical nodes.

2. The digital optimization method for trade processes according to claim 1, wherein The sparse feature vector set includes the range of transaction behavior features, the optimized value of feature screening rules, and the features associated with the correlation between transaction behavior and logistics nodes. The global critical node trend prediction value includes the dynamic change trend of core transaction nodes, the standard value of critical node lag time, and the change of connectivity weights of core transaction nodes. The distributed critical node dynamic weight matrix includes fine - grained node time - change data, time - series correlation metric values, time - change rate values, and connectivity weights of adjacent nodes. The node - associated risk transfer metrics include the combined value of risk factors between nodes, the risk conduction probability value, the abnormal change region of the weight matrix, risk - significant nodes, and the dynamic weight difference between nodes. The set of optimized paths for critical nodes includes the optimized path sequence for risk nodes, the traffic values of critical paths, weight values, and traffic distribution ratios.

3. The digital optimization method for trade processes according to claim 2, wherein, The specific steps for obtaining the distributed critical node dynamic weight matrix are as follows: S301: Based on the global critical node trend prediction value, refine the time - change data of each node, extract the fine - grained time - change of multiple nodes through time - series decomposition, analyze the change trend in combination with time - point information, and generate fine - grained node time - change data. S302: Conduct time - series correlation measurement on the fine - grained node time - change data. Calculate the time - correlation coefficient between adjacent nodes based on the time series of nodes, and select nodes with strong correlation through correlation analysis to generate the time - series correlation metric value for each node. S303: Extract the time - change rate from the time - series correlation metric values. Calculate the change rate of the node time series through the difference algorithm, screen and mark nodes with large rate differences, and generate the time - change rate value. S304: Compare the time - change rate value with the connectivity weight value of adjacent nodes, using the formula: ； Call the node update rule, adjust the weights node by node, and establish a distributed critical node dynamic weight matrix. Among them, represents the updated node weight, represents the original node weight, is the adjustment coefficient, which is used to control the influence of rate change on weight adjustment, represents the rate difference between nodes, is the average value of node rates.

4. The digital optimization method for trade processes according to claim 3, characterized in that, The specific steps for obtaining the node - associated risk transfer metrics are as follows: S401: Based on the distributed critical node dynamic weight matrix, extract the combined value of risk factors between nodes. Analyze the dynamic change rate of multiple nodes in the weight matrix, calculate the risk correlation between multiple nodes in combination with the influence value of risk factors, and generate the combined value of risk factors between nodes. S402: Calculate the conduction probability for the combined value of risk factors between nodes. Select node groups through risk correlation, analyze the transfer path and probability distribution of risk factors between nodes, and calculate the risk conduction probability in combination with the cumulative transfer model to generate the risk conduction probability value between nodes. S403: Compare and analyze the risk conduction probability value between nodes with the abnormal change region of the weight matrix, calculate the correlation between the risk conduction probability and weight change, using the formula: ； Weight the risk conduction probability in combination with the weight change rate, and screen risk - significant nodes. Among them, represents the correlation value of the significant risk node, is the conduction probability of the th node, is the weight value of the th node, is the change in node weight, is the average node weight, is the adjustment factor to control the impact of weight change; S404: Analyze the dynamic weight differences of the risk-significant nodes. By comparing the matching degree between the dynamic weight changes among nodes and the conduction probability values, combine the risk correlation degree to calculate the risk transfer relationship among nodes, and establish the node-associated risk transfer index.

5. The digital optimization method for trade processes according to claim 4, wherein The specific steps for obtaining the set of optimized paths for the key nodes are as follows: S501: Based on the node-associated risk transfer index, through per-node analysis of the risk conduction probability and weight change values among nodes, screen the key nodes with relatively high conduction probability among the risk nodes, and combine the path flow characteristics to determine the initial optimized paths, generating an optimized path sequence for the risk nodes; S502: Adjust the critical path flow values and weight values of the optimized path sequence of the risk nodes. Calculate the contribution weights of multiple nodes based on the proportion of node flow within the path, reallocate the flow values through the total weight difference rate, calculate the adjusted path connectivity index, and adjust the critical path flow values and weight values; S503: Verify the connectivity of the optimized paths after adjusting the critical path flow values and weight values. By calculating the impact of the flow distribution ratio on the path connectivity, use the formula: ； Optimize the standardized distribution of the node flow on the path, generate the optimized flow distribution in combination with the weight adjustment ratio, and adjust the flow distribution ratio; Among them, represents the path connectivity index, which measures the influence of the traffic distribution and weights among the nodes in the path. is the traffic value of the -th node on the path, indicating the transmission capacity of this node. is the weight value of the corresponding node, indicating the importance of the node. is the total number of nodes on the path, providing a reference for the path scale. is the adjustment factor, which is used to control the influence of traffic fluctuations on path connectivity. represents the standard deviation of the path node traffic, measuring the uniformity of the node traffic distribution; S504: Conduct an analysis of the optimization characteristics of the paths after adjusting the flow distribution ratio. By calculating the flow balance degree and node connectivity matching rate within the path, screen the paths according to the balanced flow distribution and connectivity, and generate the set of optimized paths for the key nodes according to the optimization rules.

6. A digital optimization system for trade processes, characterized in that, According to the trade process digital optimization method according to any one of claims 1-5, the system includes: The data feature extraction module extracts the time-series feature values of the transaction orders, the lag time values of the logistics nodes, and the abnormal transaction behavior amounts based on the trade data set, calculates the differences of multiple feature values, and calculates the range of the transaction behavior features, compares and adjusts the feature screening rules, optimizes the associated features of the transaction behavior and the logistics nodes, and generates a set of sparse feature vectors; The global trend analysis module calculates the global trade data change trend sequence values based on the set of sparse feature vectors, extracts the dynamic change trends of the core transaction nodes, determines whether the lag time values of the key nodes meet the standards, analyzes the changes in the connectivity weights of the core transaction nodes, and verifies the consistency, establishing the global key node trend prediction values; The key node optimization module separates the fine-grained node time change data based on the global key node trend prediction values, calculates the time-series correlation metric values of each node, extracts the time change rate values, compares the connectivity weight values of adjacent nodes, and calls the node update rules to iteratively optimize the node association, establishing a distributed key node dynamic weight matrix; The risk factor analysis module extracts the risk factor combination values among nodes based on the distributed key node dynamic weight matrix, calculates the risk conduction probability values, analyzes the abnormal change regions in the weight matrix, screens the risk-significant nodes, compares the dynamic weight differences among nodes, and establishes the node-associated risk transfer index; Based on the node-associated risk transfer index, the path optimization module constructs an optimized path sequence for risk nodes, adjusts the flow value and weight value of the critical path, verifies the connectivity of the optimized path, adjusts the flow distribution ratio, and generates a set of optimized paths for critical nodes.

Images