A supply chain enterprise data authenticity analysis method

Abstract

The scheme discloses a kind of supply chain enterprise data authenticity analysis method, comprising: S1. obtaining N original data of enterprise with supply chain upstream and downstream relationship;S2. respectively extract key data field in N original data of enterprise, obtain N enterprise data set;S3. respectively extract corresponding enterprise sub data set from N enterprise data set, obtain at least N enterprise sub data set;S4. the enterprise sub data set of mutual correlation is obtained by merging two by two target database;S5. each enterprise sub data set obtains feature vector based on corresponding target database;S6. the similarity between the data feature vectors corresponding to the enterprise sub data set of mutual correlation is calculated;S7. data authenticity is judged according to similarity result.The scheme proposes authenticity analysis for the data used for enterprise health degree analysis, which can ensure the effectiveness and authenticity of subsequent enterprise health degree analysis.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data analysis, and particularly relates to a supply chain enterprise data authenticity analysis method. BACKGROUND

[0002] Enterprise data generally refers to information and data related to enterprise operation, including but not limited to product information and operation data. The supply chain refers to a functional network structure connecting suppliers, manufacturers, distributors and end users into an integral whole, starting from supporting parts, manufacturing intermediate products and final products, and finally delivering products to consumers by a sales network.

[0003] Under the background of the supply chain, in order to ensure the normal operation of the enterprise itself and reduce the cooperation cost and risk of upstream and downstream enterprises, enterprise data needs to be analyzed in a timely and efficient manner to analyze the health degree of the target enterprise industry ecology. In addition, in the financial industry, it is also necessary to verify the enterprise ecological data, which can enable the creditor (generally a bank) to more clearly grasp the health degree of the enterprise ecology of the loan applicant. At present, all the health degree analysis of the target enterprise is based on the analysis of existing data, and does not focus on the authenticity of the data. However, in the process of analyzing the health degree of the enterprise based on the data, the authenticity of the enterprise data actually plays a key role in the analysis result. Only by ensuring the authenticity of the analyzed data can the effectiveness of the subsequent analysis based on the data be ensured. SUMMARY

[0004] The purpose of the present application is to provide a supply chain enterprise data authenticity analysis method to solve the above problems.

[0005] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0006] A supply chain enterprise data authenticity analysis method comprises the following steps:

[0007] S1. Obtain N original data of enterprises having upstream and downstream relationships in the supply chain; the data collected is the original data in a preset time period, such as the original data in the last 30 days.

[0008] S2. Extract the key data fields in the original data of the N enterprises respectively to obtain N enterprise data sets;

[0009] S3. Extract the corresponding enterprise sub-data sets from the N enterprise data sets respectively to obtain at least N enterprise sub-data sets;

[0010] S4. Merge the enterprise sub-data sets associated with each other to obtain a target database;

[0011] S5. Each enterprise sub-data set is based on the corresponding target database to obtain a feature vector;

[0012] S6. Calculate the similarity between the data feature vectors corresponding to the interrelated enterprise sub-data sets;

[0013] S7. Make a data authenticity judgment according to the similarity result.

[0014] In the above supply chain enterprise data authenticity analysis method, in step S1, according to the analysis purpose, N enterprises with upstream and downstream relationships on the supply chain are determined;

[0015] In step S2, according to the analysis purpose, the key data fields in the original data of the N enterprises are extracted, and N enterprise data sets are obtained.

[0016] In the above supply chain enterprise data authenticity analysis method, the analysis purpose includes enterprise data credibility analysis of one or more target enterprises on the supply chain;

[0017] In step S1, the N enterprises include the target enterprise and the upstream enterprises of the target enterprise and the downstream enterprises of the target enterprise; the target enterprise can have no upstream enterprise or downstream enterprise, and the upstream enterprises of the target enterprise can have multiple, and the downstream enterprises of the target enterprise can also have multiple.

[0018] In step S2, for the target enterprise, its sales data and purchase data are the key data fields; for the upstream enterprises of the target enterprise, its sales data is the key data field, and for the downstream enterprises of the target enterprise, its purchase data is the key data field.

[0019] In the above supply chain enterprise data authenticity analysis method, in step S3, according to the feature correlation between the enterprise data set of the target enterprise and the remaining enterprise data sets, the enterprise sub-data set of each enterprise is extracted;

[0020] Each enterprise sub-data set includes at least one sub-matrix.

[0021] In the above supply chain enterprise data authenticity analysis method, the enterprise sub-data set of each enterprise is as follows:

[0022] The enterprise sub-data set of the upstream enterprise of the target enterprise is a sales list set related to the target enterprise; note that the target enterprise is not limited to one.

[0023] The enterprise sub-data set of the downstream enterprise of the target enterprise is a purchase list set related to the target enterprise;

[0024] The enterprise sub-data set of the target enterprise is a purchase list set and a sales list set related to its upstream and downstream enterprises.

[0025] In step S4, the enterprise subsets of the target enterprise are merged with the enterprise subsets of each upstream and downstream enterprise in pairs.

[0026] In the above-mentioned method for analyzing the authenticity of supply chain enterprise data, in step S4, before merging the enterprise subsets in pairs, the features in each enterprise subset are merged, deduplicated, and accumulated.

[0027] In the above-mentioned method for analyzing the authenticity of supply chain enterprise data, step S5 obtains the feature vector of each enterprise subset in the following way:

[0028] When the i-th element in the target database exists in the enterprise subset and the difference in element values ​​is within the deviation threshold range, the value of the i-th element in the corresponding enterprise subset is 1; when the i-th element in the target database does not exist in the enterprise subset or the difference in element values ​​is outside the deviation threshold range, the value of the i-th element in the corresponding enterprise subset is 0.

[0029] The feature vector of the corresponding enterprise subset is obtained by acquiring the values ​​of all elements of the enterprise subset using the above method. It can be understood that the target database for calculating the feature vector of each enterprise subset is the target database resulting from merging it with another enterprise subset.

[0030] In the aforementioned method for analyzing the authenticity of supply chain enterprise data, the deviation threshold is 5%-20% of the values ​​of a smaller number of elements. This is because some deviations may occur during data entry for various reasons, and these deviations are permissible. Therefore, a deviation threshold is set, and the difference between two values ​​within this threshold is considered equal.

[0031] Alternatively, the deviation threshold is 0, meaning that when the i-th element in the target database exists in the enterprise subset and the element value is equal, the value of the i-th element in the corresponding enterprise subset is 1; when the i-th element in the target database does not exist in the enterprise subset or the element value is not equal, the value of the i-th element in the corresponding enterprise subset is 0.

[0032] In the above-mentioned method for analyzing the authenticity of supply chain enterprise data, the similarity calculation method in step S6 is as follows:

[0033] , This represents the number of elements at the same position in the feature vectors of two related enterprise subsets, where all elements result in 1. This represents the number of elements at the same position in the feature vectors corresponding to two related enterprise subsets where at least one is 1.

[0034] In the above-mentioned method for analyzing the authenticity of supply chain enterprise data, in step S7, the credibility of the target enterprise's data is analyzed based on the similarity results of the target enterprise and the preset enterprise data analysis rules.

[0035] The advantages of this invention are:

[0036] 1) Propose conducting authenticity analysis on the data used for enterprise health status analysis to ensure the effectiveness and authenticity of subsequent enterprise health status analysis;

[0037] 2) By leveraging the fact that data among supply chain companies is often consistent in certain aspects, the authenticity of data can be judged by comparing the degree of correlation between data from two or more companies. This allows for a reliable, credible, and relatively accurate analysis of the authenticity of company data.

[0038] 3) By constructing a target database to compare the degree of correlation between enterprise data, we can avoid the adverse effects of high data dimensionality, complex data format, and large data volume on analysis. Attached Figure Description

[0039] Figure 1 This is a flowchart of the supply chain enterprise data authenticity analysis method of the present invention;

[0040] Figure 2 This is a data processing flowchart using three companies as examples in the supply chain enterprise data authenticity analysis method of the present invention;

[0041] Figure 3 This is a schematic diagram illustrating the merging of two enterprise subsets into a target database in the supply chain enterprise data authenticity analysis method of the present invention;

[0042] Figure 4 This is a schematic diagram illustrating the feature vectors obtained from two enterprise subsets based on a target database in the supply chain enterprise data authenticity analysis method of the present invention. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0044] To promote the healthy development of supply chain enterprises, this solution provides a method for analyzing the authenticity of supply chain enterprise data. This method provides demanders, such as investors and banks, with the analysis results of enterprise data authenticity, offering them data analysis basis for important decision-making and eliminating the financial risks caused by enterprise data fraud.

[0045] like Figure 1 As shown, this method specifically includes the following steps:

[0046] S1. Obtain raw data from the ERP systems or other systems of various enterprises that have upstream and downstream supply chain relationships. Here, we set it to obtain raw data from N enterprises. Of course, this data should have a time period, which can be preset by the user in the system. For example, we can retrieve the raw data of the most recent 30 days each time.

[0047] S2. Extract the key data fields from the original data of N enterprises respectively to obtain N enterprise datasets;

[0048] S3. Extract corresponding enterprise subsets from N enterprise datasets to obtain at least N enterprise subsets, each of which includes at least one submatrix;

[0049] S4. Merge the interrelated enterprise subsets in pairs to obtain the target database; interrelated means, for example, the sales data of enterprise X to enterprise Y and the purchase data of enterprise Y to enterprise X. Here, the sales data and purchase data are interrelated data.

[0050] S5. Each enterprise subset obtains a feature vector based on the corresponding target database;

[0051] S6. Calculate the similarity between the data feature vectors corresponding to interrelated enterprise subsets;

[0052] S7. Determine the authenticity of the data based on the similarity results.

[0053] Specifically, in step S1, N enterprises with upstream and downstream relationships in the supply chain are identified based on the analysis objective. The analysis objective includes the credibility analysis of enterprise data of one or more target enterprises in the supply chain. Users can request credibility analysis of one or more target enterprises in the supply chain within the system; this request constitutes the analysis objective. The N enterprises include the target enterprise, its upstream enterprises, and its downstream enterprises. There can be one, two, or more target enterprises. A target enterprise may not have any upstream or downstream enterprises, and it may have multiple upstream and downstream enterprises.

[0054] In step S2, key data fields are extracted from the original data of N enterprises according to the analysis objectives, resulting in N enterprise datasets. Specifically, for the target enterprise, its sales data and procurement data are the key data fields; for the upstream enterprise of the target enterprise, its sales data are the key data fields; and for the downstream enterprise of the target enterprise, its procurement data are the key data fields.

[0055] In step S3, enterprise subsets of each enterprise are extracted based on the feature correlation between the target enterprise's enterprise dataset and the datasets of other enterprises. For example, the sales data of enterprise X to enterprise Y and the purchase data of enterprise Y to enterprise X are data with feature correlation. The sales data of enterprise X to enterprise Y can be extracted as an enterprise subset of enterprise X, and the purchase data of enterprise Y to enterprise X can be extracted as an enterprise subset of enterprise Y. Subsequently, these two subsets can be merged into the target database for similarity calculation between enterprises X and Y.

[0056] In summary, the enterprise subsets for each company are as follows:

[0057] The upstream enterprise subset of the target enterprise is a set of sales lists related to the target enterprise;

[0058] The downstream enterprise dataset of the target enterprise is a set of purchase lists related to the target enterprise;

[0059] The target company's enterprise subset consists of purchase lists and sales lists related to its upstream and downstream enterprises.

[0060] The number of enterprise subsets for non-target enterprises is the same as the number of target enterprises. When there is only one target enterprise, there is only one enterprise subset for non-target enterprises; when there are two target enterprises, there are two enterprise subsets for non-target enterprises, and so on. Similarly, the number of enterprise subsets for target enterprises is related to the number of upstream and downstream enterprises they have.

[0061] Specifically, upstream and downstream companies determine whether they have a characteristic association with the target company based on information in the sales and purchase lists, such as customer names, supplier names, specifications, and contract numbers. Similarly, the target company also determines whether it has a characteristic association with upstream and downstream companies based on the aforementioned information.

[0062] In step S4, before merging the enterprise subsets pairwise, the features in each enterprise subset are merged and de-overlapped. For example, if an upstream enterprise subset contains two sales lists involving 100 units of component A and 200 units of component A, the two features can be merged to remove duplicates of the same feature. Then, the two feature values ​​are superimposed to obtain 300 units of component A. Then, each subset of the target enterprise is merged pairwise with the relevant enterprise subsets of each upstream and downstream enterprise.

[0063] Specifically, in step S5, the feature vector of each enterprise subset is obtained in the following way:

[0064] When the i-th element in the target database exists in the enterprise subset and the difference in element values ​​is within the deviation threshold range, the value of the i-th element in the corresponding enterprise subset is 1; when the i-th element in the target database does not exist in the enterprise subset or the difference in element values ​​is outside the deviation threshold range, the value of the i-th element in the corresponding enterprise subset is 0.

[0065] The feature vector of the corresponding enterprise subset is obtained by acquiring the values ​​of all elements of the enterprise subset using the above method.

[0066] The deviation threshold is 5%-20% of the values ​​of a few elements, such as 10%, because some deviations may occur during data entry for various reasons, and these deviations are allowed. Therefore, a deviation threshold is set, and the difference between the two values ​​is considered to be within the deviation threshold range.

[0067] Of course, the deviation threshold can also be 0. That is, when the i-th element in the target database exists in the enterprise subset and the element value is equal, the value of the i-th element in the corresponding enterprise subset is 1. When the i-th element in the target database does not exist in the enterprise subset or the element value is not equal, the value of the i-th element in the corresponding enterprise subset is 0.

[0068] Specifically, in step S6, the similarity is calculated as follows:

[0069] , This represents the number of elements at the same position in the feature vectors of two related enterprise subsets, where all elements result in 1. This represents the number of elements at the same position in the feature vectors corresponding to two related enterprise subsets where at least one is 1.

[0070] In step S7, the authenticity of the target company's data is analyzed based on the similarity results.

[0071] Data credibility analysis can be obtained based on the similarity results and preset enterprise data analysis rules. The preset enterprise data analysis rules can be:

[0072] Calculate the average similarity of all the target company, such as P = (P1 + P2 ... + Pn) / n, where P1, P2 ... Pn represent the similarity between the target company and each of its upstream and downstream companies, and P represents the average similarity.

[0073] Several threshold ranges are set, such as 0 to V1, V1 to V2, ... and V4 to V5, each corresponding to a certain degree of data correlation for a company. For example: an average similarity of 0 to V1 indicates low data credibility; an average similarity of V1 to V2 indicates relatively low data credibility; an average similarity of V2 to V3 indicates moderate data credibility; an average similarity of V3 to V4 indicates relatively high data credibility; and an average similarity of V4 to V5 indicates extremely high data credibility.

[0074] Of course, other analysis rules can also be used. Specific enterprise data analysis rules are preset by those skilled in the art based on actual conditions, and will not be restricted or elaborated here.

[0075] Preferably, the credibility analysis results can be presented in the form of an analysis report, which includes the similarity between the target company and its upstream and downstream companies, as well as the final credibility conclusion.

[0076] Furthermore, this solution also provides a verification method:

[0077] When a user requests result verification, the data reliability of the upstream and downstream companies involved in the target company is verified. Each upstream and downstream company is treated as the target company in turn, and the reliability is calculated in the above way. If the reliability is higher than the set threshold, the data reliability result verification is passed. Otherwise, the upstream and downstream companies with reliability lower than the set threshold are removed from the supply chain relationship, and the reliability of the target company's data is reassessed.

[0078] Furthermore, this solution also provides methods for verifying supply chain companies:

[0079] When a user requests enterprise verification, the system provides the user with a list of upstream and downstream companies involved in the target enterprise for selection. The system then calculates the reliability of one or more of the user's selected upstream and downstream companies sequentially as target enterprises, using the methods described above, and outputs the reliability results to the user. Based on the reliability results, the user can choose to confirm the verification passing or customize the removal of one or more upstream and downstream companies to re-analyze the authenticity of the target enterprise.

[0080] By excluding upstream and downstream companies whose data reliability is questionable through the above methods, the accuracy and reliability of the authenticity analysis can be further ensured.

[0081] Furthermore, this solution also provides customized methods for supply chain companies:

[0082] After a user selects a target company, the system displays several upstream and downstream companies that have a relationship with the target company. The user can choose to have all upstream and downstream companies participate in the data authenticity analysis of the target company, or select only some upstream and downstream companies, or select all upstream and downstream companies after excluding some. This combination of human intervention and system integration improves the accuracy of the final analysis results.

[0083] like Figure 2 As shown, to better understand this solution, we take three companies with upstream and downstream supply chain relationships—Company X, Company Y, and Company Z—as an example. Company X is a component supplier, Company Y is a chip manufacturer, and Company Z is an equipment manufacturer. Company X supplies components to Company Y, and Company Y supplies chips to Company Z. Of course, in practical applications, more companies will be involved, and their businesses may overlap. A company is not limited to one type of product; for example, Company Z could also be a chip manufacturer. For ease of understanding, this is simplified here. In practical applications, the method provided can be expanded accordingly. In this example, the analysis aims to conduct data credibility analysis with Company Y as the target company.

[0084] The sales and purchase data of company Y are extracted as key data fields. Simultaneously, the sales data of its upstream company (company X in this case) and the purchase data of its downstream company (company Z in this case) are extracted as key data fields. This results in three company datasets: X, Y, and Z, denoted as DX, DY, and DZ, respectively. Then, based on the characteristic relationships between company Y and companies X and Z, sub-datasets are extracted for each company. Specifically, sales data for company Y are extracted from DX (denoted as DX1), purchase data for company Y is extracted from DZ (denoted as DZ1), and purchase data for company X and sales data for company Z are extracted from DY (denoted as DY1 and DY2, respectively). Since there are typically multiple business transactions between two companies (e.g., company Y purchases from company X three times), there will be three sales / purchase lists related to company X and company Y. Therefore, DX1, DY1, DY2, and DZ1 each have multiple sub-matrices. Since the same components are often involved in each procurement, in order to process the data more efficiently, this solution merges multiple sub-matrices to remove overlap and perform addition processing. The component features in each enterprise subset will only involve one value.

[0085] Merging DX1 and DY1 yields target database one; merging DY2 and DZ1 yields target database two.

[0086] Based on DX1 and the target database, obtain the data feature vector PX1 corresponding to DX1;

[0087] Based on DY1 and the target database, obtain the data feature vector PY1 corresponding to DY1;

[0088] Based on DY2 and target database two, obtain the data feature vector PY2 corresponding to DY2;

[0089] Based on DZ1 and target database 2, the data feature vector PZ1 corresponding to DZ1 is obtained.

[0090] like Figure 3 As shown, taking the two subsets DX1 and DY1 as examples for further explanation, DX1 has three sub-matrices after deduplication, involving 9 components, and DY1 has two sub-matrices after deduplication, involving 9 components. The two subsets are merged to obtain the target database one involving 18 elements.

[0091] like Figure 4 As shown, Figure 4 When the i-th element in the target database exists in DX1 and the element values ​​are equal, i is... Figure 4 In the dataset, 1, 2, 3... 18 corresponds to the i-th element of DX1 having a value of 1. When the i-th element in the target database does not exist in DX1 or has a different value, the i-th element of the corresponding enterprise subset has a value of 0. This is how the dataset is obtained. Figure 4 In the PX1 subset, when the i-th element in the target database exists in DY1 and its value is equal, the value of the i-th element in DY1 is 1; when the i-th element in the target database does not exist in DY1 or its value is unequal, the value of the i-th element in the corresponding enterprise subset is 0. This process is repeated. Figure 4 From PY1, we can obtain S=10 / 18. This similarity is the similarity between enterprise Y and enterprise X, denoted as SXY.

[0092] Finally, the similarity scores SXY and SZY between the target company Y and companies X and Z are obtained. Based on SXY, SZY and the preset enterprise data analysis rules, the final credibility analysis results can be given.

[0093] The above examples are three companies using a single chain. In actual operation, there will be many more companies.

[0094] Of course, if a situation exists where the target company has a subset of enterprise data relative to company K, but company K does not have a subset of enterprise data relative to the target company, then the correlation between the target company and company K is set to 0. This extreme case is generally unlikely to occur.

[0095] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for analyzing the authenticity of supply chain enterprise data, characterized in that, Includes the following steps: S1. Obtain raw data from N companies with upstream and downstream supply chain relationships. The N companies include the target company and its upstream and downstream companies. S2. Extract the key data fields from the original data of N enterprises respectively to obtain N enterprise datasets; For the target company, its sales data and procurement data are key data fields; for the target company's upstream companies, its sales data is a key data field; and for the target company's downstream companies, its procurement data is a key data field. S3. Extract enterprise subsets for each enterprise based on the feature correlations between the target enterprise's enterprise dataset and the datasets of other enterprises, resulting in at least N enterprise subsets; Among them, the upstream enterprise subset of the target enterprise is a set of sales lists related to the target enterprise; The downstream enterprise dataset of the target enterprise is a set of purchase lists related to the target enterprise; The target company's enterprise subset consists of purchase lists and sales lists related to its upstream and downstream enterprises; S4. Merge each enterprise subset of the target enterprise with the enterprise subsets related to each upstream and downstream enterprise to obtain the target database; S5. Obtain the values ​​of all elements in the enterprise subset to obtain the feature vector of each enterprise subset using the following method: When the i-th element in the target database exists in the enterprise subset and the difference in element values ​​is within the deviation threshold range, the value of the i-th element in the corresponding enterprise subset is 1; when the i-th element in the target database does not exist in the enterprise subset or the difference in element values ​​is outside the deviation threshold range, the value of the i-th element in the corresponding enterprise subset is 0. S6. Calculate the similarity between the data feature vectors corresponding to interrelated enterprise subsets; S7. Analyze the data credibility of the target company based on the similarity results and preset enterprise data analysis rules, including: Calculate the average similarity of all the target companies, P = (P1 + P2 ... + Pn) / n, where P1, P2 ... Pn represent the similarity between the target company and each of its upstream and downstream companies, and P represents the average similarity. Several threshold intervals are set, each corresponding to the degree of correlation of enterprise data. The data credibility of the target enterprise is determined based on the threshold interval in which the average similarity value is located.

2. The method for analyzing the authenticity of supply chain enterprise data according to claim 1, characterized in that, In step S1, N enterprises with upstream and downstream relationships in the supply chain are identified according to the analysis objectives; In step S2, key data fields are extracted from the original data of N enterprises according to the analysis purpose to obtain N enterprise datasets.

3. The method for analyzing the authenticity of supply chain enterprise data according to claim 2, characterized in that, The purpose of the analysis includes the credibility analysis of corporate data of one or more target companies in the supply chain.

4. The method for analyzing the authenticity of supply chain enterprise data according to claim 3, characterized in that, In step S4, before merging the enterprise subsets pairwise, the features in each enterprise subset are merged, deduplicated, and accumulated.

5. The method for analyzing the authenticity of supply chain enterprise data according to claim 4, characterized in that, The aforementioned deviation threshold is 5%-20% of the smaller value of the same element in the subsets of two related enterprises; Alternatively, the deviation threshold is 0, meaning that when the i-th element in the target database exists in the enterprise subset and the element value is equal, the value of the i-th element in the corresponding enterprise subset is 1; when the i-th element in the target database does not exist in the enterprise subset or the element value is not equal, the value of the i-th element in the corresponding enterprise subset is 0.

6. The method for analyzing the authenticity of supply chain enterprise data according to claim 5, characterized in that, In step S6, the similarity is calculated as follows: , This represents the number of elements at the same position in the feature vectors of two related enterprise subsets, where all elements result in 1. This represents the number of elements at the same position in the feature vectors corresponding to two related enterprise subsets where at least one is 1.

Images