Intelligent supply chain financial service platform

Abstract

The invention relates to the technical field of financial service management, in particular to an intelligent supply chain financial service platform, and the platform comprises a supply chain data analysis module which receives and integrates market dynamics, carries out the format standardization and data cleaning of the market dynamics, and generates a data integration result; and performing dimension reduction on the data integration result by applying principal component analysis, and extracting factors influencing the health degree of the supply chain to obtain a key factor extraction result. According to the invention, standardized integration and data cleaning are carried out on market dynamics, reliability and consistency of supply chain data are enhanced, data dimension reduction is realized in cooperation with principal component analysis, key factors having the most influence on the health state of the supply chain are extracted, and accuracy and sensitivity of supply chain state evaluation are improved; the risk features of the supply chain entity are analyzed by adopting a clustering algorithm, behavior pattern analysis is performed on the basis of risk feature grouping, potential risk points are found in advance, and the risk pre-judgment capability and prevention and control accuracy are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of financial service management technology, and in particular to a smart supply chain financial service platform. Background Art

[0002] The field of financial services management technology refers to a comprehensive technical field that uses information technology, big data analysis, artificial intelligence, blockchain, the Internet of Things and other advanced technologies to optimize, manage and support decision-making for activities such as the allocation of financial resources, risk control, operation of financial business, customer relationship management, and innovation of financial products and services.

[0003] Existing technologies mostly focus on the application of simple information technology and the static configuration of financial resources. They lack in-depth exploration of factors such as the dynamic characteristics of the supply chain, potential risks, and transparency of capital flows. Data collection and processing have the drawbacks of extensive management. Market dynamic data has not been effectively standardized, making it difficult to form accurate analysis conclusions. Therefore, improvements are needed. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose an intelligent supply chain financial service platform.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solutions: The smart supply chain financial service platform includes: The supply chain data analysis module receives and integrates market dynamics, standardizes the format of market dynamics and cleans the data to generate data integration results; applies principal component analysis to reduce the dimension of the data integration results, extracts factors that affect the health of the supply chain, and obtains key factor extraction results; The risk identification and clustering module uses a clustering algorithm to analyze the risk characteristics of entities in the supply chain based on the data integration results, clusters and generates risk characteristic groups of various entities, and obtains risk characteristic grouping results; based on the risk characteristic grouping results, analyzes the behavior patterns of each risk group, identifies potential risk points, and generates potential risk point analysis results; The financial transparency evaluation module extracts the capital turnover speed and loan transparency based on the key factor extraction results, performs variance analysis, compares the transparency between different time periods and financial products, and obtains the comprehensive evaluation results of financial transparency; The prevention measures and optimization module designs targeted risk prevention measures and implements risk mitigation strategies based on the potential risk point analysis results to obtain risk prevention strategy design results; and uses the comprehensive financial transparency evaluation results to adjust the operating procedures of supply chain financial services to obtain optimized supply chain financial service results.

[0006] Preferably, the steps of obtaining the data integration result are: Collect market dynamic information released by various channels through web crawlers, including news reports, market reports and social media dynamics, filter the collected information, remove data from non-target markets, and then unify the date format, currency unit and language conversion to obtain formatted market dynamics; Based on the formatted market dynamics, duplicate records are deleted, erroneous data entries are corrected, and missing values ​​are filled, including cleaning fields with abnormal prices and missing supply chain sources to form cleaned market dynamics; Based on the cleaned market dynamics, the data is aggregated through entity recognition to obtain a data integration result.

[0007] Preferably, the steps of obtaining the key factor extraction results are: Based on the data integration results, the sales, inventory and logistics data of the supply chain nodes are obtained, the covariance matrix of the data is calculated by principal component analysis, the principal components of the data are extracted by eigenvalue decomposition, and the data set after dimensionality reduction is obtained; Based on the reduced-dimensional data set, the contribution rate of each principal component is analyzed, and the contribution rate of each principal component is sorted and screened to obtain screened key factor data; Based on the screened key factor data, regression analysis is used to verify the influence of key factors and form key factor extraction results.

[0008] Preferably, the steps of obtaining the risk feature grouping result are: Based on the data integration results, extract entity information in the supply chain, screen enterprises, logistics nodes and capital circulation links with transaction records, financial data and cooperation information, build a risk feature data set including transaction frequency, transaction amount, capital turnover rate and contract performance, and generate entity risk feature data; According to the entity risk characteristic data, the degree of deviation of the entity risk characteristics is calculated, and the calculation formula is: ; in, For Entity The degree of risk characteristic deviation Representing Entity In the The value of the risk feature, represents the total number of risk characteristics, Representative The mean of the risk characteristics in all entities, Representative The maximum value of the risk characteristics, Representative The minimum value of a risk feature, Representing Entity Total value across all risk characteristics; Based on the degree of deviation of the risk characteristics, a risk category classification standard is constructed, and entities are classified according to the degree of deviation of the risk characteristics to obtain a risk characteristic grouping result.

[0009] Preferably, the steps for obtaining the potential risk point analysis results are: Based on the risk feature grouping results, the behavior pattern abnormality value is calculated using the following formula: ; in, Risk Grouping The behavior pattern is outlier, Risk Grouping The transaction frequency in the transaction time dimension, is the median transaction frequency of all risk groups, Risk Grouping Deviations from internal contract performance, is the median contract performance deviation of all risk groups, Risk Grouping The number of abnormal transactions within Risk Grouping The number of internal fund turnovers, and They are the maximum and minimum transaction frequencies of all risk groups respectively; Based on the behavioral pattern outliers, risk groups are screened, violations or default behaviors are matched, and whether there are similarities with historical behaviors is analyzed to generate potential risk point analysis results.

[0010] Preferably, the steps for obtaining the comprehensive evaluation results of financial transparency are: Based on the extraction results of the key factors, extract the capital turnover speed and loan transparency data of financial products in different time periods to generate a preliminary data set of financial transparency; Based on the preliminary data set of financial transparency, the transparency index of each financial product is calculated using the following formula: ; in, For financial products The transparency index, For financial products The speed of loan approval, For financial products The default rate, For financial products The speed of capital turnover, For financial products The efficiency of capital recovery; Based on the transparency index, we compare the transparency of different time periods and different financial products, analyze the fluctuation trend of the transparency index, evaluate the overall transparency health of the financial market, and obtain a comprehensive evaluation result of financial transparency.

[0011] Preferably, the steps for obtaining the risk prevention strategy design result are: According to the analysis results of the potential risk points, the risk impact index is calculated using the following formula: ; in, Risk Source The risk impact index of Representative risk sources The abnormal capital flow value, Represents the average abnormal fund flow of all risk sources, Representative risk sources The number of credit defaults, Representative risk sources Deviations in contract execution, Representative risk sources The frequency of abnormal transactions in the past period; Based on the risk impact index, the risk mitigation strategy is matched, and credit freezing, transaction restrictions, dynamic monitoring or additional review mechanisms are adopted for risk sources of different levels to obtain the risk prevention strategy design results.

[0012] Preferably, the steps for obtaining the optimized supply chain financial service results are: Based on the comprehensive evaluation results of financial transparency, identify the time delays in loan approval, the opacity of fund allocation and the inadequacy of risk management measures in the financial service process, sort out the factors affecting process efficiency, and generate a list of key actions for improvement; Based on the key action list, redesign the supply chain financial service process and establish optimized supply chain financial service results.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the reliability and consistency of supply chain data are enhanced by standardizing and integrating market dynamics and cleaning data, data dimension reduction is achieved in conjunction with principal component analysis, the key factors that have the greatest impact on the health status of the supply chain are extracted, and the accuracy and sensitivity of the supply chain status assessment are improved; a clustering algorithm is used to analyze the risk characteristics of supply chain entities, and behavioral pattern analysis is performed based on risk characteristic grouping to discover potential risk points in advance, thereby enhancing the ability to predict risks and the accuracy of risk prevention and control; based on the variance analysis of capital turnover speed and loan transparency, a differentiated transparency assessment of different financial products and time periods is achieved, thereby improving the transparency of supply chain financial services and the efficiency of capital utilization; risk prevention measures are designed for potential risk points, risk mitigation strategies are optimized, risk events are avoided in a timely manner, and operational risks are reduced; the financial service operation process is adjusted in a timely manner based on the evaluation results of financial transparency, the allocation and operation of supply chain capital resources are realized, the stability and flexibility of the overall operation of the supply chain are strengthened, and the financing efficiency of the enterprise and the market competitiveness of the supply chain are comprehensively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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 intended to limit the present invention.

[0016] See also Figure 1 The present invention provides a technical solution: the smart supply chain financial service platform includes: The supply chain data analysis module receives and integrates market dynamics, standardizes the format of market dynamics and cleans the data to generate data integration results; it applies principal component analysis to reduce the dimension of data integration results, extracts factors that affect the health of the supply chain, and obtains key factor extraction results; The risk identification and clustering module uses clustering algorithms to analyze the risk characteristics of entities in the supply chain based on the data integration results, clusters and generates risk characteristic groups for various entities, and obtains risk characteristic grouping results; based on the risk characteristic grouping results, conducts behavior pattern analysis for each risk group, identifies potential risk points, and generates potential risk point analysis results; The financial transparency evaluation module extracts the capital turnover speed and loan transparency based on the key factor extraction results, performs variance analysis, compares the transparency between different time periods and financial products, and obtains the comprehensive evaluation results of financial transparency; The preventive measures and optimization module designs targeted risk prevention measures and implements risk mitigation strategies based on the results of potential risk point analysis to obtain risk prevention strategy design results; it uses the comprehensive evaluation results of financial transparency to adjust the operating procedures of supply chain financial services to obtain optimized supply chain financial service results.

[0017] The steps to obtain the data integration results are: Collect market dynamic information released by various channels through web crawlers, including news reports, market reports and social media dynamics, filter the collected information, remove data from non-target markets, and then unify the date format, currency unit and language conversion to obtain formatted market dynamics; Based on the formatted market dynamics, duplicate records are deleted, incorrect data entries are corrected, and missing values ​​are filled, including cleaning of fields with abnormal prices and missing supply chain sources to form cleaned market dynamics; Based on the cleaned market dynamics, the data is aggregated through entity recognition to obtain data integration results.

[0018] Specifically, when accessing the target website and social platform through a web crawler, first list the searchable keyword combinations in the crawling script and set the matching range for them, such as using "supply chain finance", "market fluctuations" and "logistics dynamics" as search conditions, and record the source, release time and topic label of each search result. If the topic label does not match the pre-established industry label list, the record will be regarded as non-target market information and discarded. The industry label list is manually compiled based on previous market research data and contains about 30 fields and key industry names, such as retail, manufacturing, agriculture, etc. Then, the date format of the retained news reports and market reports is converted, and all records that are not in the "YYYY-MM-DD" format are uniformly rewritten as "YYY The official exchange rate summary table is updated daily from public financial data sources. Assuming that the base currency is RMB, it can be converted uniformly according to the exchange rate of RMB against USD, EUR, GBP and JPY on that day. In the language conversion link, if the input text is detected to be in English, French or Japanese, the trained neural network translation model is called to translate the whole sentence. The training data of the model includes about 30,000 translation pairs of common financial vocabulary and supply chain proprietary terms. The learning rate is specified to be 0.001 during training, and 20 rounds are iterated. After the error converges, a relatively stable translation output can be obtained. Finally, all processed texts are merged and classified by source to obtain formatted market dynamics.

[0019] Based on the formatted market dynamics, it is first checked for duplicates according to the unique index. The unique index here is composed of the source identifier, release time and main product name. If the index is completely consistent, it is considered a duplicate record and deleted. Then the value range of the price field is compared one by one, with the span of 0 to 1,000,000 as the basic judgment, and the average value of the surrounding price fluctuation within 5% of the historical statistical data as a reference. If the price value of a single record exceeds this range, it is marked as abnormal and manually compared. When manually comparing, it is set that if similar records do not have similar extreme values ​​in the same time period If the value is incorrect, you can choose to interpolate and correct it based on the average price of the surrounding time periods. For the missing supply chain source field, if there are more than 15 missing data entries (the 15 entries come from the empirical statistics of the crawling scale and loss rate in the previous three months), a detailed backtracking is initiated by retrieving the original page or trying to query relevant information from other sources in the same time period. If the completion still fails, it is directly marked as a null value and filtered in subsequent processing. Finally, all records that have executed price corrections, completed missing fillings, or been retained after judgment are summarized to form the market dynamics after cleaning.

[0020] Based on the cleaned market dynamics, the data needs to be aggregated through entity recognition. First, the fields such as the company name, logistics node and capital flow link in each record are annotated, and the corresponding entity tags are found in the text. These key information are identified through the pre-trained sequence annotation model. The model is trained using a bidirectional long short-term memory network plus a conditional random field structure. About 10,000 manually annotated news texts and social media posts are used as training samples. The number of training rounds is set to 30 rounds, and the batch size of each round is 64. The initial learning rate is 0.002 and decays to half of the original after every 10 rounds. After the model converges, it can automatically identify information such as the full name of the company and the logistics node code in the text. Then the identified entities are classified and summarized, and all records of the same company or logistics node are aggregated according to date, transaction type, and capital amount. The entity name plus time range is used as the primary index. Then, a series of numerical fields (such as capital investment, transaction amount, etc.) of the same time period and the same entity are concentrated into one record by summing or taking the average. Finally, all aggregated records are unified and sorted to obtain data integration results.

[0021] The steps to obtain the key factor extraction results are: Based on the data integration results, the sales, inventory and logistics data of the supply chain nodes are obtained, the covariance matrix of the data is calculated by principal component analysis, and the principal components of the data are extracted by eigenvalue decomposition to obtain the reduced-dimensional data set; Based on the reduced-dimensional data set, the contribution rate of each principal component is analyzed, and the contribution rate of each principal component is sorted and screened to obtain the screened key factor data; Based on the screened key factor data, regression analysis is used to verify the influence of key factors and form the key factor extraction results.

[0022] Specifically, based on the data integration results, we first read the sales quantity, inventory and logistics turnover records of each supply chain node, and assemble all the records into a matrix form, where the rows represent data points in different time periods or batches, and the columns represent items such as sales, inventory days and transportation intervals. This matrix is ​​recorded as And calculate its mean vector Then, construct the covariance matrix The specific calculation method of the covariance matrix is ,in is the number of records, It is For data vectors of records, if the value of an indicator is too large, it can be compared with the historical monitoring range first. For example, if the sales volume exceeds 2,000,000, it is necessary to further verify whether there are similar records maintained at a similar level, and then perform eigenvalue decomposition on the above covariance matrix. The specific calculation of eigenvalues ​​and eigenvectors can be completed by the determinant method or the numerical iteration method. If an eigenvalue is less than 0.05, the minimum threshold set empirically, it is considered that the information corresponding to the feature accounts for too low a proportion and can be ignored. The 0.05 is estimated based on the recent supply chain scale and sales level, and is verified by the test set before each iteration to confirm that the coverage of the principal component information is relatively complete under this value. The remaining eigenvectors and eigenvalues ​​are arranged from large to small. After intercepting several vectors with higher eigenvalues, the reduced-dimensional data set is obtained by multiplying the vector with the original data matrix.

[0023] Based on the reduced-dimensional data set, we first obtain the contribution rate of each principal component and form a list. The contribution rate refers to the proportion of the variance of a principal component to the overall variance. The principal components can be expressed as If you want to set the screening threshold of the principal component contribution rate, you can set the total cumulative contribution rate to 80% or the single principal component contribution rate to be no less than 5%. These two figures are based on the business scope and data dimensions of the supply chain node in the past two years, and are adjusted in combination with the field test data. By comparing the contribution rate of each principal component with the above threshold, if the cumulative contribution rate is in the range of 80% to 90% and the single principal component contribution rate is greater than or equal to 5%, these principal components will be included in the subsequent processing. For the case where the contribution rate exceeds 90% but the single principal component contribution rate is too low, it is necessary to re-check whether its numerical source is normal, such as checking whether its explanatory power in sales volume or inventory changes is consistent with the statistical results of the most recent month. Finally, a group of data vectors corresponding to the principal components that meet the screening rules are extracted and recorded as the key factor data after screening.

[0024] Based on the screened key factor data, regression analysis is needed to verify the influence of these key factors. First, a regression model is constructed for each key factor and an objective function is set. For example, sales profit is used as the dependent variable, and inventory turnover, transportation frequency and the corresponding principal component scores are used as independent variables. The regression coefficient is estimated by ordinary least squares or other common fitting methods. If ordinary least squares is used, it can be defined as ,in is the observed profit value, The fitting value is calculated based on the current regression coefficient, and then the regression model is trained in combination with a certain number of known samples. For each iteration, the coefficient convergence status needs to be recorded and it is determined whether the preset error threshold is met. For example, the error can be set between 5000 and 10000 and the shape of the fitting curve can be observed. If the coefficient diverges or the error cannot be reduced below the threshold, it is necessary to review the abnormal points of the input data again. After completing the convergence calculation of all regression coefficients, the influence of each key factor on the sales profit can be summarized and output as the key factor extraction result.

[0025] The steps to obtain the risk feature grouping results are as follows: Based on the data integration results, we extract entity information in the supply chain, screen enterprises, logistics nodes and capital circulation links with transaction records, financial data and cooperation information, build a risk feature data set including transaction frequency, transaction amount, capital turnover rate and contract performance, and generate entity risk feature data; According to the entity risk characteristic data, the degree of deviation of the entity's risk characteristics is calculated using the following formula: ; in, For Entity The degree of risk characteristic deviation Representing Entity In the The value of the risk feature, represents the total number of risk characteristics, Representative The mean of the risk characteristics in all entities, Representative The maximum value of the risk characteristics, Representative The minimum value of a risk feature, Representing Entity Total value across all risk characteristics; Based on the degree of risk characteristic deviation, a risk category classification standard is constructed, and entities are classified according to the degree of risk characteristic deviation to obtain the risk characteristic grouping results.

[0026] Specifically, based on the previously acquired data integration results, first read the enterprises and logistics nodes registered in all supply chain nodes, call the corresponding data source for the financial information of each entity, and extract numerical fields such as transaction frequency, transaction amount, capital turnover rate, and contract performance within a quarter, and compare these values ​​with the valid ranges listed in advance. For example, limit the transaction frequency to between 1 and 200 times per month, limit the capital turnover rate to between 0 and 1, and limit the transaction amount to between 0 and 1,000,000. For records that exceed this range, immediately trigger data tracing and extract the data from the account transaction information. The actual data is verified. For contract performance, the number of executed contracts in the past six months is compared with the number of defaults or extensions. When the default ratio is greater than 10%, the record is marked as high risk. This 10% is obtained based on the default statistics of similar scales and industries in the past year and is confirmed through multiple batch operations and analysis. Subsequently, the values ​​of various indicators are aggregated according to the unique identifier of the enterprise or logistics node. After completing the integrity check of the aggregated data, it is formally summarized into a risk feature data set that includes transaction frequency, transaction amount, capital turnover rate, and contract performance, and generates entity risk feature data.

[0027] The benefit of the formula is that it introduces a cubic measure of skewness and a comprehensive calculation of the ratio of range to total value, taking into account the possible asymmetric distribution of various risk characteristics in actual business scenarios and differences of different orders of magnitude.

[0028] The steps to obtain the risk feature data set are: find the first Risk characteristics in the entity The observed values ​​at the transaction, including transaction amount, transaction frequency, capital turnover rate or contract fulfillment ratio, need to be obtained through clear quantitative methods. For example, the transaction amount can be summarized from all fund transaction records in a period of time, the transaction frequency needs to count the total number of transactions completed by the entity in that period, the capital turnover rate can be subdivided and calculated based on the relationship between working capital and operating cycle, and if the contract fulfillment ratio is to be quantified, the ratio of the number of normally fulfilled contracts to the total number of contracts can be calculated. After recording all these subdivided calculations, the , if we take entity "E101" and transaction frequency as an example, the transaction frequency can be directly presented in the system as an integer. If 65 transactions were recorded in the past quarter, then , when all features have completed data collection, all Numeric value.

[0029] The steps to obtain are: calculate the first The average value of a risk feature is obtained by first extracting the values ​​of all entities on the feature into a column, then summing them up and dividing them by the total number of entities to get the mean value of the feature. The first risk feature is the transaction amount, and the corresponding transaction amount is obtained from the 100 sampled companies and recorded as , can be calculated using the following mathematical expression : When a specific example is needed, the transaction amounts of enterprise numbers E101 to E200 can be added up one by one to get 6,000,000, and then divided by the number of enterprises 100 to get 60,000.

[0030] The steps to obtain are: count the total number of risk features, count all the features listed when constructing the risk feature dataset and get an integer value. If the current analysis includes the four features of transaction frequency, transaction amount, capital turnover rate and contract performance, then .

[0031] The steps to obtain are: The maximum value is extracted from the values ​​presented on all risk features. Here, the comparison of the same dimension is required according to the numerical meaning of each feature. If the dimensions between features are not uniform, they need to be normalized first, and then the maximum value is calculated. For example, if there is a transaction frequency for entity E101 , transaction amount , capital turnover rate , contract performance ratio , the transaction amount is regarded as the value of 200000, and after distinguishing the dimensions with factors such as frequency, it can be uniformly mapped in the range of 0 to 1, obtained by Min-Max normalization, and then the capital turnover rate and contract performance ratio are kept in this range. By comparing the four values ​​after normalization, we can get If the normalized values ​​of these four numbers are 0.065, 0.9, 0.4, and 0.8, respectively, then the maximum value is 0.9 and is recorded as .

[0032] The steps to obtain are: The minimum value is selected from all the characteristic values, transaction frequency corresponds to 0.065, transaction amount corresponds to 0.9, capital turnover rate corresponds to 0.4, and contract fulfillment ratio corresponds to 0.8. After comparing these four values, the minimum value 0.065 is selected, confirming that this value belongs to the dimension corresponding to transaction frequency, so =0.065.

[0033] The steps to obtain the entity are: Here The values ​​of the features are summed. If in the example , and the four eigenvalues ​​obtained after normalization for the E101 entity are 0.065, 0.9, 0.4 and 0.8 respectively. The sum value is .

[0034] Calculation process: Taking entity E101 as an example, it is known that , after normalization, we get , assuming that across all entities, the average of these four features to If they are 0.05, 0.6, 0.45, and 0.7 respectively, we can calculate: ; Divide by And after taking the absolute value: ; Then take the cube root of the result: ; Then calculate the numerator , calculate the denominator , dividing the two gives , and finally add the above two results together: ; The results show that after normalization, the risk characteristic deviation value of entity E101 relative to other comparison objects is about 0.4451. If it is necessary to compare with other entities in the subsequent stages, the same calculation method can be used to obtain the risk characteristic deviation of each entity. And make a horizontal comparison. If If the value is significantly larger among a group of entities, it means that the entity has a higher relative difference in transaction frequency, capital turnover or contract performance. When associated with the step results, entity E101 can be included in the queue with a higher deviation level.

[0035] Based on the degree of risk characteristic deviation, first obtain the Read the deviation results of all entities in the value list and assign a unique number to each entity for tracking. Compared with the pre-defined interval range, for example, 0 to 0.3 is defined as the low deviation interval, 0.3 to 0.6 is defined as the medium deviation interval, and greater than 0.6 is defined as the high deviation interval. These interval boundaries are deduced from historical monitoring data and summarized in combination with the deviation levels of major entities in the past year. After the setting, if an entity appears If the value of falls into the high deviation range, it will be classified as a high-risk category, the medium deviation range will be recorded as a medium-risk category, and the low deviation range will be recorded as a low-risk category. After the data of all entities are compared, a detailed list with classification labels is formed, which includes the entity number, Deviation values ​​and corresponding risk category labels. For high deviation categories, further analysis can be carried out in subsequent stages to complete the risk feature grouping results.

[0036] The steps to obtain the potential risk point analysis results are as follows: Based on the risk feature grouping results, the behavior pattern abnormality is calculated using the following formula: ; in, Risk Grouping The behavior pattern is outlier, Risk Grouping The transaction frequency in the transaction time dimension, is the median transaction frequency of all risk groups, Risk Grouping Deviations from internal contract performance, is the median contract performance deviation of all risk groups, Risk Grouping The number of abnormal transactions within Risk Grouping The number of internal fund turnovers, and They are the maximum and minimum transaction frequencies of all risk groups respectively; Based on behavioral pattern outliers, risk groups are screened, violations or default behaviors are matched, and historical behavioral similarities are analyzed to generate potential risk point analysis results.

[0037] Specifically, the formula is beneficial in that it combines three different forms of difference analysis, namely, measuring the difference between transaction frequency and the median through cube root, measuring the degree of contract performance deviation in combination with nonlinear power, and measuring transaction abnormalities using the ratio of the number of abnormal transactions to the number of capital turnovers. The difference between the maximum and minimum global transaction frequencies is then introduced into the denominator. This allows the degree of abnormality of each group to be measured from multiple dimensions while taking into account differences in the overall trading environment.

[0038] parameter Steps to obtain: First find the group in the risk feature grouping results All entities or records involved, and then the transaction behavior of the group in the transaction time dimension is segmented by week or month, and the number of transactions in each segment is added up to get the total transaction frequency of the group. , it is necessary to extract the occurrence time of each transaction from the transaction data and count it. If a group contains several enterprises or logistics nodes, it is necessary to summarize all their transaction counts in the same time period. The summary number is The original data can then be used to measure the frequency in units of one quarter or half a year, depending on the length of the statistical period. If the period is too short, it is easy to cause excessive fluctuations, and if the period is too long, it will reduce the sensitivity to recent anomalies. In practice, combined with the time period of the company's quarterly or semi-annual financial report, a 3 to 6 month period is often selected for statistics. For example, for group "G5", all transactions involved in the past quarter are counted. After querying the transaction records, the total number of transactions is 220, that is, .

[0039] parameter The steps to obtain each risk group are as follows: Then, according to the group index Arrange the transaction frequencies of all groups in a column and select the median as The calculation of the median can be discussed in different cases: if the number of risk groups is odd, choose the value in the middle; if the number of risk groups is even, take the average of the two middle values. The transaction frequency data all come from the defined risk group list. Each group contains at least one corporate entity. When the number is small, the number of transactions can be directly sorted in ascending order and the position can be found. When the number is large, it can be summarized in blocks and then sorted. For a certain statistics, there are 10 risk groups. The transaction frequencies after ascending order are [80, 120, 150, 180, 220, 230, 250, 310, 350, 400]. The median is between the 5th and 6th values ​​220 and 230. Calculate , thus we get .

[0040] parameter To quantify the contract performance deviation, it is necessary to group All contract execution records involved in the group are counted, and events such as overdue, breach of contract, and early termination are extracted and compared with the total number of contracts in the group. Specifically, a deviation ratio value or deviation score can be set to reflect the contract performance. For example, if there are 30 contracts, 5 of which are overdue or have content changes to varying degrees, a scoring process can be set up first to combine the number of overdue days, overdue amount, and number of changes into a comprehensive score to obtain The value of the contract deviation can be calculated by the formula: ,in For the adjustment coefficient between the number of overdue days and the number of behavioral changes, a suitable value needs to be determined through historical large sample statistics to ensure that different overdue days and default types can be reasonably combined. Finally, the obtained results are normalized in the interval [0, 1] to facilitate comparison with other groups. For example, for group "G5", the total number of overdue days in its 30 contracts is 70 days, and the number of behavioral changes is 3. hour, .

[0041] parameter The acquisition steps are as follows: Then, sort the risk groups in ascending order and find the median of the performance deviation in the entire risk grouping set. The median is calculated in the same way. If there are several groups The data is arranged as [0.5, 0.8, 1.0, 1.1, 1.7, 2.0, 2.2, 2.5, 2.9], etc., then the corresponding distribution of the middle position can be directly retrieved. If the total number of groups is 9, then the 5th data is the median, and we get .

[0042] parameter The acquisition steps: The number of abnormal transactions can be directly in the group The screening criteria may include a single transaction amount exceeding a specified threshold, a cumulative surge in transaction frequency in a very short period of time, or unknown whereabouts of funds. Each record that meets the abnormality judgment criteria will be recorded as an abnormal transaction. In the management process, an amount or number threshold is often formulated in combination with the company's financial review system, and is set based on the average level of similar companies in the past three years. For example, a single transaction amount greater than 500,000 yuan is recorded as abnormal, or more than 10 consecutive transactions in a single day are also recorded as abnormal. Then, each transaction is compared within the group range, and finally accumulated. For example, in a group G5, there are 8 transactions with a transaction amount of more than 500,000 yuan in the past month, and 4 transactions occurred intensively in a short period of time on the same day. After the two are combined and deduplicated, a total of 10 transactions are considered abnormal records. Therefore .

[0043] parameter Steps to obtain: Funds turnover times represent grouping To understand the cumulative process of funds flowing from one party to another, it is necessary to read all the financial transaction details of the group in a certain time period. Each transfer or fund transfer between accounts can be registered as a circulation. If a group brings together multiple enterprises or logistics nodes, all transfers generated by them should be included to obtain a total circulation number. In order to avoid a large value due to a long interval, it can be synchronized with the accounting cycle of the financial statement. For example, re-counting is performed every quarter. If it is found that a group has 85 fund transfers in three months, then it will be recorded as .

[0044] parameter Steps to obtain: transaction frequency values ​​for all risk groups List them into a set and compare them, extracting the highest value to determine For example, there are 10 risk groups with transaction frequencies of [80, 120, 150, 180, 220, 230, 250, 310, 350, 400], where the maximum value is 400. .

[0045] parameter Steps to obtain: Similarly, the transaction frequencies of all groups After the values ​​are concentrated, the minimum value is extracted. , still taking the above data as an example, the lowest value is 80, .

[0046] Calculation process: Here we take group G5 as an example to perform calculations. It is known that: ; And across all risk groups, , ; The first step is to calculate : ; ; The second step is to calculate : ; After numerical calculation, it can be approximated as: ; Step 3: Calculate : ; Add the above three parts together: ; Then the denominator ; So the whole formula calculates the result: ; The results show that for group G5, the deviation between the transaction frequency and the median is relatively controllable. At the same time, the contract performance deviation and capital flow also show a certain degree of abnormality, but the overall value is around 0.00751, which means that within the range of all risk groups, the abnormal value of G5's behavior pattern is in a relatively low range. If it is necessary to compare with other groups, the formula can be used to calculate and obtain their respective , and then combine it with certain interval divisions to determine which group has larger outliers or is more worthy of attention.

[0047] Based on the behavior pattern outliers, first retrieve the group details of all completed calculations The values ​​are then indexed and the transaction frequency, contract performance deviation, abnormal transaction number, fund flow number and other factors are listed for each group according to its unique identifier. The value is compared with the preset threshold, where the preset threshold is set mainly by combining the average abnormal grouping level in the past year and superimposing the fluctuation coefficient. If the values ​​are all within the range of 0.01 to 0.05, 0.05 can be set as the threshold and adjusted up or down appropriately according to business needs. For example, it can be set to 0.06 to maintain a certain monitoring width for potential anomalies. This threshold can also be dynamically adjusted with quarterly updates during system configuration. Then check the group list one by one. If there is If it exceeds 0.06, it will be marked as a more serious anomaly and its transaction and financial details will be reviewed. By comparing the actual violations or breaches in the group with historical archives, especially checking whether there are similar surges in the number of transactions, significant increases in contract performance deviations, or a high number of abnormal transactions in previous years or quarters, high-probability repeated risk signs can be identified. If the current group is found to be similar to previous records, a specific symbol will be added to the list for prompting. All comparison and annotation information will be collected in an integrated table to obtain the potential risk point analysis results.

[0048] The steps to obtain the comprehensive evaluation results of financial transparency are as follows: Based on the key factor extraction results, extract the capital turnover speed and loan transparency data of financial products in different time periods to generate a preliminary data set of financial transparency; Based on the preliminary data set of financial transparency, the transparency index of each financial product is calculated using the following formula: ; in, For financial products The transparency index, For financial products The speed of loan approval, For financial products The default rate, For financial products The speed of capital turnover, For financial products The efficiency of capital recovery; Based on the transparency index, we compare the transparency of different time periods and different financial products, analyze the fluctuation trend of the transparency index, evaluate the overall transparency health of the financial market, and obtain a comprehensive evaluation result of financial transparency.

[0049] Specifically, based on the key factor extraction results obtained above, first select the target product's capital turnover speed index and loan transparency index one by one from the archived financial product list, summarize all transaction flow records and approval process records of each product in the past two quarters, and then extract them in segments according to the transaction date and approval date as the time axis, and read the total capital turnover time and the average approval feedback time for each segment, divide the total capital turnover time by the number of capital transactions in the corresponding period to obtain the value of the phased capital turnover speed, divide the difference between the approval feedback time and the issuance time by the number of transactions submitted for approval to obtain the basic data on loan transparency, and if it is found that data for certain periods are missing or there is data delay, the system will be updated for that product. The corresponding financial transactions or risk control links are traced back to verify whether there is a cross-platform or cross-institutional approval process, and all data are checked against a pre-established reasonable range. For example, the capital turnover speed is set in the range of 1 day to 30 days. If a product exceeds 30 days, it is listed as an abnormal item and the approval reason is tracked again. The 30-day range is obtained by adding the average loan period of financial products in the statistics of the past year and the elastic time limit. If the loan transparency data is lower than a certain range, it will be recorded in subsequent links. The range can be set between 0.5 and 1.0, which is determined by observing the average feedback level of similar products under strict review conditions in the past. Finally, the capital turnover speed and loan transparency of each time period are summarized to form a preliminary data set of financial transparency.

[0050] The benefit of the formula lies in that it comprehensively measures the transparency of financial products from four aspects: loan approval speed, default rate, capital turnover speed, and capital recovery efficiency. It achieves nonlinear balance through logarithmic and fourth root operations, and can find the combination point between indicators of different orders of magnitude that takes into account both amplification and reduction effects.

[0051] parameter Steps to obtain: This parameter represents the financial product The loan approval speed is mainly based on recording the time of initiation and completion of each loan application in the system background, and then calculating the average approval time by counting the adjacent time differences. If the loan approval is cross-departmental or cross-institutional, the processing time of these links is additionally combined, and the arithmetic mean of all samples is taken to form the final In addition, extreme values ​​will be checked, for example, if an approval is delayed too long or is artificially suspended, it will be excluded from the analysis to ensure that the Basically reflects the conventional approval efficiency of the product. For example, a financial product F1 has initiated 120 loan applications in the past three months. The total time from application to approval is 840 days. Days is recorded as the average approval time, and the sky.

[0052] parameter Steps to obtain: This parameter represents the financial product The default rate requires retrieving the number of defaults or the amount of arrears from all the contracted loans of the financial product and quantifying them. Usually, the repayment status of each loan is recorded in the financial system, and the default rate is generated by counting all the default records within a certain period and calculating the proportion. , for example, the formula or , if a total of 400 loans were signed for the F1 product in the past year, and 25 of them were overdue for more than 30 days, .

[0053] parameter Steps to obtain: This parameter represents the financial product To determine the capital turnover speed, it is necessary to quantify the frequency of loan issuance and payment collection during the operating cycle of the product. The method is to count the interval between the loan issuance time point and the payment collection time point, and combine the amount of funds to obtain the average turnover speed. If a payment is completed 10 days after the loan issuance, the number of days occupied by this amount of funds is 10 days. The average number of days occupied by funds of all samples can be obtained , the total amount of funds occupied can also be weighted according to the business scale. If the average number of days for funds occupied in the F1 product is 15 days based on the 500 records in the last six months, then .

[0054] parameter Steps to obtain: This parameter is used to measure the financial product The fund recovery efficiency is as follows: For example, in the first quarter, the maturity amount of a product F1 is RMB 6 million, of which RMB 4 million is recovered on time on the maturity day or the next day, RMB 1 million is recovered within one week after maturity, and another RMB 1 million is recovered more than 30 days later. The weight of on-time recovery can be set to 1.0, the weight of short-term delayed recovery to 0.4, and the weight of seriously overdue recovery to 0.1. The specific value can be obtained by multiplying the recovery amount by the corresponding weight, summing it up and dividing it by the total maturity amount: .

[0055] Calculation process: Take the F1 financial product as an example to perform a complete calculation and obtain its parameters in advance: ; The first step is to calculate the numerator : ; ; ; Step 2: Calculate the denominator : ; ; ; ; Calculate the quadratic root first: , and then take the second root of 15.03: , get ; Step 3, divide the numerator by the denominator: ; The results show that the comprehensive transparency index of the financial product F1 in terms of approval speed, default, capital flow and recovery efficiency is approximately 0.4866. If a judgment threshold is set in advance, 0.4 to 0.6 can be regarded as a medium transparency range, then F1 falls within this range and its relative position can be further determined by comparing it with the transparency index of other products.

[0056] Based on the transparency index, we first select the transparency index of financial products that have been calculated in the past three quarters, and mark different time period labels for each product to ensure that the data of the first quarter, second quarter and third quarter can be distinguished, and then the corresponding time period of each quarter is The values ​​are defined as separate sequences one by one, and these values ​​are compared to see if there is a significant increase or decrease in different quarters. The transparency index values ​​of each product in the three quarters are subtracted or divided one by one, and the index change range is recorded. In order to facilitate screening, a fluctuation threshold, such as 0.15, can be set. This threshold comes from the historical fluctuation level of the transparency index of all products in the previous year and the assessment scope of the financial regulatory report. If the transparency index of a product differs by more than 0.15 in adjacent quarters, it will be listed as an abnormal concern object. At the same time, the distribution of values ​​between different products in the same quarter should be compared vertically. If the transparency index of some products falls below 0.2, it will be listed as a low transparency group, if it falls between 0.2 and 0.6, it will be listed as a medium transparency group, and if it is greater than 0.6, it will be listed as a high transparency group. These specific intervals are adjusted with reference to the measurement results of the past two years and combined with the industry statistics of average approval time and capital recovery efficiency. After that, the quarterly series and group labels of all products are sorted out to identify whether there is a long-term low transparency or drastic fluctuation, and the summarized analysis results are marked in the overall evaluation list of the financial market as a comprehensive evaluation result of financial transparency.

[0057] The steps to obtain the risk prevention strategy design results are: According to the analysis results of potential risk points, the risk impact index is calculated using the following formula: ; in, Risk Source The risk impact index of Representative risk sources The abnormal capital flow value, Represents the average abnormal fund flow of all risk sources, Representative risk sources The number of credit defaults, Representative risk sources Deviations in contract execution, Representative risk sources The frequency of abnormal transactions in the past period; Based on the risk impact index, the risk mitigation strategy is matched, and credit freezing, transaction restrictions, dynamic monitoring or additional review mechanisms are adopted for risk sources of different levels to obtain the risk prevention strategy design results.

[0058] Specifically, the benefit of the formula lies in that it makes a comprehensive calculation of three aspects: the 1.5-power difference between the abnormal capital flow value and its mean, the square root of the sum of the squares of the number of credit defaults and the contract execution deviation, and the logarithmic penalty term of the abnormal transaction frequency. It can balance the differences in capital flow and credit execution of different risk sources in the same indicator. At the same time, extremely large or extremely small values ​​can be suppressed or amplified in layers through logarithmic terms, so that high and medium anomalies of risk sources can be distinguished.

[0059] parameter Steps to obtain: This parameter represents the risk source To obtain the abnormal fund flow value, it is necessary to count suspicious or large amount of liquidity data one by one within the specified time window, and then normalize or unify these values ​​and combine them to obtain a comprehensive indicator. , you can follow the following process: First query the risk source in the business system For all corresponding transaction flows, set "amount thresholds" or "suspicious flags" according to internal financial regulations, and accumulate the inflows and outflows of funds that meet the thresholds or have suspicious characteristics. If there are multiple statistical intervals at the same time, you can first distinguish the statistics and then merge and average them. For example, if the "S101" company retrieves a total of 4 million yuan of suspected abnormal inflows and outflows in a quarter and confirms that the total inflow and outflow scale is 80 million yuan, you can use Calculate the abnormal proportion of funds.

[0060] parameter The steps to obtain: This parameter represents the average abnormal capital flow of all risk sources. Perform summing or averaging operations. ,in is the number of risk sources, and each risk source has its abnormal capital flow value calculated in the same time window. In order to eliminate the interference of extreme data, extreme abnormal points are often removed in advance or processed by using the median plus a certain correction factor. However, as long as the method is consistent, it is also comparable in the final classification comparison. For example, 50 risk sources are identified in a time period, and these risk sources are calculated separately. Then add up all the values ​​to get 300, and divide it by 50. .

[0061] parameter Steps to obtain: This parameter indicates the risk source The number of credit defaults needs to be retrieved and counted in the business system or external credit reporting system. There are usually clear criteria for determining whether it is a credit default, such as the company's failure to repay the money on time or the violation of the contract terms causing related legal disputes. If the amount weight or the severity of the default is introduced, the number of times can also be weighted by an additional formula. For example, a company "S101" has failed to perform its credit contracts on time three times in the past six months, two of which involved delayed delivery of goods and one involved overdue loan repayment. After internal review, they are combined and counted as 3, and we get .

[0062] parameter Steps to obtain: This parameter represents the risk source The contract execution deviation is often quantified by the actual differences generated from the signing of the contract to the completion of the contract execution, which may include changes in the quality of goods, changes in the delivery date, and execution content exceeding the scope of the contract. A scoring system can be designed for quantification, such as the contract deviation score. , calculated through multiple specific fields, formula example: ,in For the The count or degree of deviation of the deviation items, is the corresponding weight. If the contract of a risk source "S101" was extended twice in the past period, each time for 3 days, and a quality deduction event occurred, then the extension event can be set =Total extension days 6 days plus one weight , add points to the quality deduction events, and finally add up all the points to generate a deviation score, which is recorded as .

[0063] parameter Steps to obtain: This parameter indicates the risk source To check the frequency of abnormal transactions in the past period, you need to retrieve all transaction records and timestamp each record. Then, when filtering data, count the transactions based on the statistical period (e.g., the past three months) and abnormal criteria. You can consider excessively frequent transactions per day, repeated large amounts of funds in and out within a short period of time, and other behaviors as abnormal transactions and list them in the Transaction Log. For example, if "S101" finds 12 suspicious transactions in a row within one month, then .

[0064] Calculation process: The following example takes "S101" as the risk source, given: The first step is to calculate the molecular part : ; ; The second step is to calculate : ; ; ; ; Add the two parts together: ; Step 3: Calculate the denominator : ; ; ; Final calculation : ; The results show that for the risk source "S101", its comprehensive risk impact index during this cycle is approximately 9.345, which is a relatively high value. It may be necessary to include more stringent control in the subsequent process or further investigate the flow of funds and credit performance.

[0065] Based on the risk impact index, first retrieve the risk impact index of each risk source obtained in the same statistical period. Then sort all risk sources by their unique codes and sort them by Different intervals can be defined by referring to historical monitoring conclusions. For example, 0 to 5 is classified as a low-risk interval, 5 to 10 is classified as a medium-risk interval, and those above 10 are classified as high-risk intervals. These numerical intervals are collected from the risk impact index distribution of the previous four quarters and determined in combination with regulatory requirements. If some risk sources If the number exceeds 10, it will be recorded as a high-risk group, and the details such as the number of credit defaults and contract execution deviations will be read to determine the specific risk inducement. Then the risk source will be compared with the previous data to confirm whether there are frequent abnormal transactions or abnormal capital flows in the recent period. If the comparison finds that the existing If the level has exceeded the medium-risk range for many times, the pre-established credit freeze or transaction restriction mechanism will be triggered, and the frequency or amount of subsequent transaction approval procedures for the risk source will be further reduced. When grading the risk sources in the medium-risk range, dynamic monitoring strategies can be set to regularly check whether the number of credit defaults continues to rise or whether the contract execution deviation has increased sharply. These strategies include weekly data collection and update and monthly on-site verification processes to record slight fluctuations in risk indicators. Regular attention is also paid to risk sources in the low-risk range. If If the abnormal capital flow value suddenly increases or the frequency of abnormal transactions significantly increases in the subsequent cycle and exceeds the established threshold, it will be automatically included in the higher-level tracking list. Finally, after integrating the processing methods and grouping information of all risk sources, the risk prevention strategy design results can be obtained.

[0066] The steps to obtain the optimized supply chain financial service results are: Based on the comprehensive evaluation results of financial transparency, identify the time delays in loan approval, the opacity of fund allocation and the inadequacy of risk management measures in the financial service process, sort out the factors affecting process efficiency, and generate a list of key actions for improvement; Based on the key action list, redesign the supply chain financial service process and establish optimized supply chain financial service results.

[0067] Specifically, based on the comprehensive evaluation results of financial transparency, first screen all recent loan approval process records and read the difference between the approval submission time and the loan completion time. If the difference exceeds the preset range of 20 to 30 days, it is marked as a time delay. The range of 20 to 30 days is determined by referring to the average approval time in the industry statistics in the past two years and combining with the compliance regulations of some banks. Then, for each loan with a delay mark, track the person in charge, approval link and number of rollbacks in the approval chain one by one, record the stagnation time at each review node and compare it with the latest processing time in the original approval process plan. If the actual time exceeds 10% of the expected time limit, it is deemed to be not completed on time. At the same time, extract the detailed data of fund allocation from all process records and check the loan and specific If the account ownership is inconsistent with the original plan or the same funds are circulated across multiple channels in a short period of time, it is identified as a case of opaque fund allocation. For this, multiple comparisons based on the transfer path can be performed in the system, and each account number passed through the transfer path can be matched with the existing corporate legal person information or related entities and included in the statistics. The registration of risk management measures for all processes and the final risk control status are then cross-checked. If the key risk review link is missing or the review opinion is unclear at the originally set approval node, it is regarded as insufficient risk management measures. Finally, the time delay, opaque fund allocation and lack of review links are summarized together, and the factors affecting the process efficiency are sorted out from the perspectives of approval flow, contract signing, repayment monitoring, etc., to generate a list of key actions for improvement.

[0068] Based on the key action list, we first reassigned the personnel responsible for approval and their functions from the loan approval node, and disassembled the original process that required multiple signatures into a method of approval in sequence and marking the approval time limit in the system. If the pre-set 48-hour approval time benchmark is exceeded, the management personnel will be immediately prompted to expedite the process. The 48 hours is the average approval time calculated after referring to the previous business processes of similar scales and daily human resource allocation. Then, the fund allocation link is set as an independent process to clarify the specific outgoing and flow destination of the funds, and automatic monitoring is set to record the repeated transfer in and out of the same amount. If it occurs within a week, If there are three or more large-scale repeated transfers, a suspicious mark will be automatically added. The three-time threshold is determined based on the weekly monitoring data in the company's internal historical statistical annual report, and a mandatory review step is added to the risk management measures, requiring that subsequent loan actions be blocked in advance when potential risks are judged to exist. For this purpose, a unique identification number is bound to each loan in the system and its approval opinions and contract signing process are tracked. Loans that do not meet the preset risk standards but whose data indicators are close to the threshold are submitted to internal auditors for one-on-one verification. Finally, the above links are arranged in a new order to form a new version of the supply chain financial service process, and an optimized supply chain financial service result is established.

Claims

1. Smart supply chain financial service platform, characterized by: The platform includes: The supply chain data analysis module receives and integrates market dynamics, standardizes the format of market dynamics and cleans the data to generate data integration results; applies principal component analysis to reduce the dimension of the data integration results, extracts factors that affect the health of the supply chain, and obtains key factor extraction results; The risk identification and clustering module uses a clustering algorithm to analyze the risk characteristics of entities in the supply chain based on the data integration results, clusters and generates risk characteristic groups of various entities, and obtains risk characteristic grouping results; based on the risk characteristic grouping results, analyzes the behavior patterns of each risk group, identifies potential risk points, and generates potential risk point analysis results; The financial transparency evaluation module extracts the capital turnover speed and loan transparency based on the key factor extraction results, performs variance analysis, compares the transparency between different time periods and financial products, and obtains the comprehensive evaluation results of financial transparency; The prevention measures and optimization module designs targeted risk prevention measures and implements risk mitigation strategies based on the potential risk point analysis results to obtain risk prevention strategy design results; and uses the comprehensive financial transparency evaluation results to adjust the operating procedures of supply chain financial services to obtain optimized supply chain financial service results.

2. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the data integration result are: Collect market dynamic information released by various channels through web crawlers, including news reports, market reports and social media dynamics, filter the collected information, remove data from non-target markets, and then unify the date format, currency unit and language conversion to obtain formatted market dynamics; Based on the formatted market dynamics, duplicate records are deleted, erroneous data entries are corrected, and missing values ​​are filled, including cleaning fields with abnormal prices and missing supply chain sources to form cleaned market dynamics; Based on the cleaned market dynamics, the data is aggregated through entity recognition to obtain a data integration result.

3. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the key factor extraction results are: Based on the data integration results, the sales, inventory and logistics data of the supply chain nodes are obtained, the covariance matrix of the data is calculated by principal component analysis, the principal components of the data are extracted by eigenvalue decomposition, and the data set after dimensionality reduction is obtained; Based on the reduced-dimensional data set, the contribution rate of each principal component is analyzed, and the contribution rate of each principal component is sorted and screened to obtain screened key factor data; Based on the screened key factor data, regression analysis is used to verify the influence of key factors and form key factor extraction results.

4. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the risk feature grouping result are: Based on the data integration results, extract entity information in the supply chain, screen enterprises, logistics nodes and capital circulation links with transaction records, financial data and cooperation information, build a risk feature data set including transaction frequency, transaction amount, capital turnover rate and contract performance, and generate entity risk feature data; According to the entity risk characteristic data, the degree of deviation of the entity risk characteristics is calculated, and the calculation formula is: ； in, For Entity The degree of risk characteristic deviation Representing Entity In the The value of the risk feature, represents the total number of risk characteristics, Representative The mean of the risk characteristics in all entities, Representative The maximum value of the risk characteristics, Representative The minimum value of a risk feature, Representing Entity Total value across all risk characteristics; Based on the degree of deviation of the risk characteristics, a risk category classification standard is constructed, and the entities are classified according to the degree of deviation of the risk characteristics to obtain the risk characteristic grouping result.

5. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the potential risk point analysis results are as follows: Based on the risk feature grouping results, the behavior pattern abnormality value is calculated using the following formula: ； in, Risk Grouping The behavior pattern is outlier, Risk Grouping The transaction frequency in the transaction time dimension, is the median transaction frequency of all risk groups, Risk Grouping Deviations from internal contract performance, is the median contract performance deviation of all risk groups, Risk Grouping The number of abnormal transactions within Risk Grouping The number of internal fund turnovers, and They are the maximum and minimum transaction frequencies of all risk groups respectively; Based on the behavioral pattern outliers, risk groups are screened, violations or default behaviors are matched, and whether there are similarities with historical behaviors is analyzed to generate potential risk point analysis results.

6. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the comprehensive evaluation results of financial transparency are as follows: Based on the extraction results of the key factors, extract the capital turnover speed and loan transparency data of financial products in different time periods to generate a preliminary data set of financial transparency; Based on the preliminary data set of financial transparency, the transparency index of each financial product is calculated using the following formula: ； in, For financial products The transparency index, For financial products The speed of loan approval, For financial products The default rate, For financial products The speed of capital turnover, For financial products The efficiency of capital recovery; Based on the transparency index, we compare the transparency of different time periods and different financial products, analyze the fluctuation trend of the transparency index, evaluate the overall transparency health of the financial market, and obtain a comprehensive evaluation result of financial transparency.

7. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the risk prevention strategy design results are: According to the analysis results of the potential risk points, the risk impact index is calculated using the following formula: ； in, Risk Source The risk impact index of Representative risk sources The abnormal capital flow value, Represents the average abnormal fund flow of all risk sources, Representative risk sources The number of credit defaults, Representative risk sources Deviations in contract execution, Representative risk sources The frequency of abnormal transactions in the past period; Based on the risk impact index, the risk mitigation strategy is matched, and credit freezing, transaction restrictions, dynamic monitoring or additional review mechanisms are adopted for risk sources of different levels to obtain the risk prevention strategy design results.

8. The smart supply chain financial service platform according to claim 1 is characterized in that: The steps for obtaining the optimized supply chain financial service results are as follows: Based on the comprehensive evaluation results of financial transparency, identify the time delays in loan approval, the opacity of fund allocation and the inadequacy of risk management measures in the financial service process, sort out the factors affecting process efficiency, and generate a list of key actions for improvement; Based on the key action list, redesign the supply chain financial service process and establish optimized supply chain financial service results.

Images