Supply chain management-oriented data structured processing and graph generation method and system
By performing detailed data preprocessing and structured process processing in the supply chain management database, combined with testing scenario optimization, the problem of low quality of knowledge graphs in the existing technology is solved, and a high-quality supply chain management knowledge graph that meets application needs is generated , providing accurate business insights and decision-making support.
Patent Information
- Application Number
- CN202411418552.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-12
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When generating knowledge graphs, the existing technology directly passes single-pretreatment, resulting in the graph structure being too simple and low in quality, which cannot meet application needs, and there is a large deviation between the utilization of the graph and the actual application results.
By collecting multimodal data in the supply chain management database, performing detailed preprocessing, including data cleaning, noise filtering and standardization, determining the quality indicators of preprocessing completion, and performing structured process processing based on the threshold to generate a supply chain management knowledge graph , and combine test scenarios for feedback and optimization.
It significantly improves the quality and consistency of multimodal data, generates a more comprehensive and accurate supply chain management knowledge graph, meets user needs, provides accurate business insights and prediction capabilities, and ensures the effective application of the graph in supply chain scenarios.
Smart Images

Figure CN118939642A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of structured data processing, and specifically to a method and system for structured data processing and graph generation for supply chain management. Background Art
[0002] As supply chain management becomes more complex, it inevitably involves a large amount of multimodal data (such as text, charts, sensor data, etc.). These data are scattered in different formats and require effective structured processing and integration to generate valuable supply chain maps.
[0003] For example, the invention patent with announcement number CN112148735B announces a method for constructing a knowledge graph for structured tabular data, including the following steps: obtaining source data; preprocessing the source data to generate a word forest; filtering the entity core from the word forest according to the entity core search algorithm; based on the entity core, according to the main table screening algorithm, determining the entity main table, entity relationship table and invalid table corresponding to the entity in turn; combining rule combination and maximum continuous character matching to determine entity attributes; determining entity relationships according to entity attributes, entity main table, entity relationship table and table foreign key constraint information to complete the construction of the ontology; according to the ontology, importing the tabular data into the graph database to obtain the corresponding knowledge graph.
[0004] For example, the invention patent with publication number CN117932089A discloses a data analysis method based on knowledge graph, which involves the field of knowledge graph technology, including: collecting data sets from multiple different sources containing structured data and unstructured data and preprocessing them; using a data integration method based on pattern matching to convert them into data sets with unified format and structure; extracting entities and relationships in the data set to establish a knowledge graph as a first knowledge graph; obtaining entity triples and relationship triples of the first knowledge graph; constructing a graph neural network; constructing a connectivity matrix to calculate the connection probability between entity triples and relationship triples and the graph neural network; according to the connection probability, connecting the first knowledge graph and the trained graph neural network to generate a probabilistic knowledge graph; and using the probabilistic knowledge graph to analyze the data set.
[0005] Combining the above technical solutions, it is found that in the current technical process of generating knowledge graphs, directly constructing the graph through unified preprocessed data will make the structure of the graph too simple. The unified preprocessing operation will increase the possibility of low graph quality, resulting in a large deviation between the results produced by subsequent use and analysis of the graph and the actual application results, causing the graph to be unable to meet application requirements. Summary of the invention
[0006] In view of the deficiencies in the prior art, the present invention provides a method and system for data structured processing and graph generation for supply chain management, which can effectively solve the problems involved in the above-mentioned background technology.
[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: The first aspect of the present invention provides a data structured processing and graph generation method for supply chain management, including: based on an established supply chain management database, collecting supply chain related data of various data modes, and performing preprocessing operations, thereby determining the preprocessing completion quality indicators of various data modes; comparing the preprocessing completion quality indicators of various data modes with preset preprocessing completion quality thresholds to obtain comparison results, and based on the comparison results, performing structured process processing to generate a supply chain management knowledge graph; through the supply chain management knowledge graph, combined with the supply chain management test scenario, testing the supply chain management knowledge graph, thereby obtaining the test results of the supply chain management knowledge graph, and providing feedback on the test of the supply chain management knowledge graph.
[0008] As a further method, the data cleaning abnormality feedback is performed, and the specific feedback process is: Based on the comparison results of cleaning values less than the preset cleaning threshold, a number of supply chain data points corresponding to cleaning values less than the preset cleaning threshold are counted and recorded as abnormal supply chain data points of various data modes; the preset cleaning threshold is subjected to difference processing with the cleaning value corresponding to each abnormal supply chain data point of each data mode to obtain the cleaning deviation value corresponding to each abnormal supply chain data point of each data mode, and the cleaning deviation value is matched with the data cleaning adjustment plan corresponding to each preset cleaning deviation value interval, thereby obtaining the data cleaning adjustment plan corresponding to each abnormal supply chain data point of each data mode; based on the data cleaning adjustment plan corresponding to each abnormal supply chain data point of each data mode, the corresponding The data cleaning adjustment plan is used to re-clean each abnormal supply chain data point of each data modality to obtain the corresponding adjustment cleaning value of each abnormal supply chain data point of each data modality, and compare it with the preset cleaning threshold. If the corresponding adjustment cleaning value of an abnormal supply chain data point of a certain data modality is less than the preset cleaning threshold, the abnormal supply chain data point of this data modality is eliminated; if the corresponding adjustment cleaning value of an abnormal supply chain data point of a certain data modality is greater than or equal to the preset cleaning threshold, the abnormal supply chain data point of this data modality is recorded as the normal adjusted supply chain data point of this data modality, thereby counting the normal adjusted supply chain data points of each data modality.
[0009] As a further method, the structured process is processed based on the comparison results to generate a supply chain management knowledge graph. The specific process is as follows: Based on the comparison results, if the comparison results show that the preprocessing completion quality indicators of various data modalities are greater than or equal to the preset preprocessing completion quality threshold, it is determined that various data modalities are directly subjected to structured process processing, and a supply chain management knowledge graph is generated; if the comparison results show that the preprocessing completion quality indicators of various data modalities are less than the preset preprocessing completion quality threshold, it is necessary to perform structured preprocessing operations on various data modalities, and perform structured process processing on various data modalities that have completed the structured preprocessing operations, thereby generating a supply chain management knowledge graph.
[0010] As a further method, the supply chain management knowledge graph is tested, specifically each test user in the supply chain management test scenario, performs a graph query test through the supply chain management knowledge graph, and collects the test query result data corresponding to each test user and the test process data of each test user.
[0011] As a further method, the test of the supply chain management knowledge graph is fed back, and the specific feedback process is: The test result of the supply chain management knowledge graph is compared with the preset test threshold. If the test result of the supply chain management knowledge graph is greater than or equal to the preset test threshold, the supply chain management knowledge graph is visualized; if the test result of the supply chain management knowledge graph is less than the preset test threshold, the test of the supply chain management knowledge graph is fed back through the feedback mechanism in the supply chain management database.
[0012] The second aspect of the present invention provides a system for data structured processing and graph generation methods for supply chain management, including: a data preprocessing module, which is used to collect supply chain-related data of various data modes based on an established supply chain management database, and perform preprocessing operations, thereby determining the preprocessing completion quality indicators of various data modes; a knowledge graph generation module, which is used to compare the preprocessing completion quality indicators of various data modes with preset preprocessing completion quality thresholds to obtain comparison results, and based on the comparison results, perform structured process processing to generate a supply chain management knowledge graph; a graph test feedback module, which is used to test the supply chain management knowledge graph through the supply chain management knowledge graph in combination with the supply chain management test scenario, thereby obtaining the test results of the supply chain management knowledge graph and providing feedback on the test of the supply chain management knowledge graph.
[0013] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention provides a data structured processing and graph generation method and system for supply chain management. First, the collected supply chain related data is preprocessed in detail, and the preprocessing completion quality indicators of various data modes are determined, which can significantly improve the quality and consistency of multimodal data. Thus, a more comprehensive and accurate supply chain management knowledge graph is generated through structured process processing. The supply chain management knowledge graph is tested by combining the supply chain management test scenario, thereby obtaining the test results of the supply chain management knowledge graph. The generated supply chain management knowledge graph meets user needs through a feedback mechanism. At the same time, the high-quality supply chain graph provides accurate business insights and prediction capabilities, which ultimately enables users to accurately understand the overall picture of the relevant supply chain and provide a basis for making accurate and comprehensive supply chain management decisions in the future.
[0014] (2) The present invention collects supply chain-related data of various data modes through a pre-established supply chain management database, and performs preprocessing operations in sequence, including data cleaning, noise filtering and standardization, thereby determining the preprocessing completion quality indicators of various data modes and comparing them with the preset preprocessing completion quality thresholds, thereby providing accurate data support for subsequent structured process processing.
[0015] (3) The present invention adopts advanced structured processing technology to generate a high-quality supply chain management knowledge graph, and combines it with supply chain management test scenarios to test the supply chain management knowledge graph, thereby obtaining a supply chain management knowledge graph that satisfies users and can provide users with comprehensive, real-time and accurate information support, ensuring that the graph is effectively applied to supply chain scenarios to improve the level of information management of the supply chain. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.
[0017] Figure 1 The figure is a schematic flow chart of the method steps of the present invention.
[0018] Figure 2 It is a schematic diagram of system module connection of the present invention. DETAILED DESCRIPTION
[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0020] This example will be explained in detail with the following scenarios: For graph construction, how to obtain multimodal data and perform structured processing is the key, that is, the multimodal data needs to be structured in detail and accurately to ensure the accuracy of the constructed graph. However, after the graph is constructed, whether it can be effectively applied to the scene is also a key issue that this application needs to consider. Therefore, this application performs detailed and accurate preprocessing of the multimodal data, and tests the constructed graph to ensure that the graph can be effectively used in the scene and provide users with high-quality graphs.
[0021] This example uses new data management technologies such as BigTable to build a large database of enterprise supply chains (i.e., the supply chain management database of this application). By collecting and preprocessing massive amounts of supply chain-related data, accurate data can be obtained under the premise of building a graph, and then structured processing (explained in detail below) is used to generate a high-quality graph; the graph is then tested in combination with scenarios (here only an example is given as a specific test result), that is, users (for the scenario of supply chain management, the users here are the personnel who use the graph in each enterprise) can enter key fields in the graph interface by selecting regions, industries, sectors, bill types, etc., and the graph will provide users with information in the form of charts, including the total number of enterprises (using bills), distribution of years of establishment of enterprises, distribution of registered capital, distribution of industries to which they belong, distribution of bill payment amounts in the past year, distribution of bill payment frequencies in the past year, distribution of bill circulation times, distribution of bill payment overdue times, and other information.
[0022] It needs to be explained that the above-mentioned supply chain-related data can be the total number of enterprises (using bills) in the quantitative data mode, the distribution of registered capital, the distribution of bill payment amounts in the past year, the distribution of bill payment frequencies in the past year, the distribution of bill circulation times, the distribution of bill payment overdue times, and of course, it also includes the qualitative data mode, such as the distribution of years of establishment of enterprises and the distribution of industries to which they belong, and other supply chain-related data under multiple data modes.
[0023] Of course, the above-mentioned users can also be core enterprises. By drawing industrial chain maps and restoring industrial chain relationships, core enterprises can understand the overall picture of the supply chain, eliminate risks and redundant links on the chain, and optimize the supply chain structure, thereby enhancing the competitiveness of the supply chain.
[0024] Therefore, the knowledge graph constructed by this application can generate a series of industry graphs for each field by researching, analyzing and exploring the data of each industry and each market segment, as well as the development status and trends, and use the graphs to clearly and accurately reflect the data of each industry and each field. It helps core enterprises to grasp the overall picture of the supply chain, grasp the enterprises at the end of the chain, and assist enterprises in making development decisions.
[0025] Reference Figure 1 As shown, the first aspect of the present invention provides a data structured processing and graph generation method for supply chain management, including: based on an established supply chain management database, collecting supply chain related data of various data modalities, and performing preprocessing operations, thereby determining the preprocessing completion quality indicators of various data modalities.
[0026] It should be noted that the supply chain management database established above is not a data storage library in the traditional sense, but a database used to construct the knowledge graph in this example. This supply chain management database includes a variety of data analysis and processing algorithms, such as data cleaning operations, noise filtering operations, etc., and of course, the corresponding feedback mechanism is also preset, that is, to provide high-quality feedback information level for the subsequent knowledge graph test results.
[0027] Specifically, the preprocessing operation refers to the preprocessing operation module of the supply chain management database preprocessing the supply chain related data of various data modes, wherein the preprocessing operation module includes data cleaning, noise filtering and standardization.
[0028] In this example, data cleaning, noise filtering and standardization are carried out in sequence. First, the data is cleaned to deal with problems such as missing values, duplicate data and inconsistent formats. After the cleaning process is completed, if the data in this example meets the predetermined quality standard (that is, greater than or equal to the preset cleaning threshold), it will enter the preprocessing step of noise filtering; that is, outliers and noise are removed. This process depends on the results of cleaning to ensure the effectiveness of noise filtering, because if there are still cleaning anomalies in the data, noise filtering may be affected; after noise filtering, the data is uniformly formatted and scaled. At this time, the data should be clean and noise-free to ensure the accuracy of standardization and ultimately obtain complete, accurate and high-quality supply chain-related data.
[0029] Furthermore, the data cleaning specifically involves cleaning the supply chain-related data of various data modalities, wherein the supply chain-related data of various data modalities include various supply chain data points of various data modalities, wherein each supply chain data point refers to multiple data of each type of data modality, collectively referred to as supply chain-related data, such as the above-mentioned distribution of years of establishment of enterprises and the distribution of industries to which they belong, namely, various supply chain data points of qualitative data modalities; after data cleaning of each supply chain data point of each data modality, a corresponding cleaning value of each supply chain data point of each data modality is generated, wherein the cleaning value refers to a cleaning score preset in the supply chain management database.
[0030] The cleaning values corresponding to each supply chain data point of each data modality are compared with the cleaning threshold preset by the graph generation management platform, where the cleaning threshold refers to the reasonable minimum cleaning standard specified after the data is cleaned. If the cleaning value corresponding to a supply chain data point of a certain data modality is less than the preset cleaning threshold, the supply chain data point of this data modality is recorded as an abnormal supply chain data point of this data modality, and data cleaning abnormality feedback is performed.
[0031] If the cleaning value corresponding to a supply chain data point of a certain type of data modality is greater than or equal to the preset cleaning threshold, then the supply chain data point of this type of data modality is recorded as the normal supply chain data point of this type of data modality. Thus, a number of supply chain data points corresponding to the cleaning values greater than or equal to the preset cleaning threshold are counted and recorded as the normal supply chain data points of each type of data modality.
[0032] Specifically, the data cleaning abnormality feedback is performed, and the specific feedback process is as follows: Based on the comparison results of cleaning values less than the preset cleaning threshold, several supply chain data points corresponding to cleaning values less than the preset cleaning threshold are counted and recorded as abnormal supply chain data points of various data modes.
[0033] The preset cleaning threshold is subjected to difference processing with the cleaning value corresponding to each abnormal supply chain data point of each data modality, and the cleaning deviation value corresponding to each abnormal supply chain data point of each data modality is obtained, which is matched with the data cleaning adjustment plan corresponding to each cleaning deviation value interval preset by the graph generation management platform, thereby obtaining the data cleaning adjustment plan corresponding to each abnormal supply chain data point of each data modality. This example specifically illustrates the above matching process: The graph generation management platform divides the cleaning deviation value interval into three intervals, namely [0, 10], (10, 90] and greater than 90, which correspond to three data cleaning adjustment plans, namely low-difficulty data cleaning adjustment plan, medium-difficulty data cleaning adjustment plan and high-difficulty data cleaning adjustment plan. If the cleaning deviation value of a certain abnormal supply chain data point of a certain data mode is 50, it is matched as a medium-difficulty data cleaning adjustment plan. Specifically, the data cleaning process corresponding to the data with a cleaning deviation value of 50 is sorted out through the supply chain management database, and the problems existing in the cleaning are found and the types of data cleaning quality problems are clarified, such as missing values, duplicate values, abnormal values, inconsistent formats, etc. Therefore, the data is cleaned again based on the existing cleaning problems, that is, the overall process of the medium-difficulty data cleaning adjustment plan is completed.
[0034] Based on the data cleaning adjustment plan corresponding to each abnormal supply chain data point of each data modality, each abnormal supply chain data point of each data modality is re-cleaned with the corresponding data cleaning adjustment plan to obtain the corresponding adjustment cleaning value of each abnormal supply chain data point of each data modality, which is compared with the preset cleaning threshold. If the corresponding adjustment cleaning value of an abnormal supply chain data point of a certain data modality is less than the preset cleaning threshold, the abnormal supply chain data point of this data modality is eliminated. The above elimination is due to the fact that the data still fails to meet the cleaning requirements after being cleaned twice, and the supply chain management database will automatically delete it.
[0035] If the adjusted cleaning value corresponding to an abnormal supply chain data point of a certain type of data modality is greater than or equal to the preset cleaning threshold, then the abnormal supply chain data point of this type of data modality is recorded as the normal adjusted supply chain data point of this type of data modality, thereby counting the normal adjusted supply chain data points of each type of data modality.
[0036] Furthermore, the determination of the preprocessing completion quality index of each data modality is specifically carried out as follows: Each normal supply chain data point of each data mode and each normal adjusted supply chain data point of each data mode are subjected to noise filtering and standardization preprocessing operations in turn. The noise filtering and standardization preprocessing operation process is similar to the above-mentioned data cleaning process, and only the preprocessing problem is different, so this example will not be explained in detail.
[0037] Therefore, based on the noise filtering and standardized preprocessing operation completion process, the preprocessing completion quality indicators of various data modes are comprehensively determined. The specific comprehensive determination process is that after the normal supply chain data points of various data modes and the normal adjusted supply chain data points of various data modes are subjected to noise filtering and standardized preprocessing operations in turn, the standardized processing values corresponding to the multiple data that finally exist (if the preprocessing process is only data cleaning, then the final preprocessing completion quality indicator of each data mode is the sum of the cleaning values corresponding to each normal supply chain data point of each data mode and the adjusted cleaning values corresponding to each normal adjusted supply chain data point of each data mode; and the preprocessing process in this example is not only data cleaning, so the final preprocessing completion quality indicator of each data mode is the standardized processing value corresponding to multiple data of each data mode), collectively referred to as the preprocessing completion quality indicators of various data modes.
[0038] In a specific embodiment, the present invention collects supply chain-related data of various data modalities through a pre-established supply chain management database, and performs preprocessing operations in sequence, including data cleaning, noise filtering, and standardization, thereby determining the preprocessing completion quality indicators of various data modalities, and comparing them with preset preprocessing completion quality thresholds, to provide accurate data support for subsequent structured process processing.
[0039] The preprocessing completion quality indicators of various data modalities are compared with the preprocessing completion quality thresholds preset by the graph generation management platform to obtain the comparison results. Based on the comparison results, structured process processing is performed to generate a supply chain management knowledge graph.
[0040] Specifically, based on the comparison results, structured process processing is performed to generate a supply chain management knowledge graph. The specific process is as follows: Based on the comparison results, if the comparison results show that the preprocessing completion quality indicators of various data modes are greater than or equal to the preset preprocessing completion quality threshold, where the preprocessing completion quality threshold represents the minimum quality value obtained after the data passes the preprocessing operation (here it refers to a minimum standard set in the data preprocessing process to ensure the validity and quality of the data. Meeting this standard means that the data can meet the requirements of subsequent construction of high-quality knowledge graphs after preprocessing operations such as cleaning, noise filtering and standardization), then it is determined that various data modes are directly processed by structured processes and a supply chain management knowledge graph is generated.
[0041] The structured process processing is as follows: Use the deep learning model (such as a multimodal neural network) stored in the supply chain management database to perform cross-modal feature extraction on the preprocessed data. For example, if the data in this example is text data, use natural language processing (NLP) technology to extract key information, such as supplier name, cargo description, etc.; then combine the self-attention mechanism stored in the supply chain management database to perform feature fusion (not explained in detail here) to generate a comprehensive feature vector, and use the named entity recognition (NER) and relationship extraction technology stored in the supply chain management database to identify key entities in the supply chain (such as suppliers, products, transportation routes, etc.) and their relationships from the fused comprehensive feature vector, thereby structuring the extracted entities and relationships into a knowledge graph and storing it in the supply chain management database.
[0042] If the comparison result shows that the preprocessing completion quality index of each data modality is less than the preset preprocessing completion quality threshold, it is necessary to perform structured preprocessing operations on each data modality, wherein the structured preprocessing operation is specifically that the supply chain management database receives another data preprocessing algorithm (this example refers to the data balancing algorithm), and performs data balancing on multiple data in a certain type or multiple types of data modalities corresponding to the preprocessing completion quality index less than the preset preprocessing completion quality threshold. Since the processing time of the above data balancing is long and the operation is complex in the supply chain management database, this example only performs this operation on multiple data in a certain type or multiple types of data modalities corresponding to the preprocessing completion quality index less than the preset preprocessing completion quality threshold (since the preprocessing operation of this application is accurate and efficient, only a small amount of data needs to be balanced after the supply chain-related data is preprocessed, so this example performs a specific data balancing operation on a certain data in a certain type of data modality, and the specific process is as follows: For example, the distribution of bill payment amounts in the past year of the quantitative data mode corresponding to the preprocessing completion quality index less than the preset preprocessing completion quality threshold is identified, and the distribution of bill payment amounts in the past year in the quantitative data mode is subjected to descriptive statistical analysis through the supply chain management database (i.e., Python (using the NumPy library) in the supply chain management database) to evaluate its distribution, thereby identifying the degree of imbalance in the distribution of bill payment amounts in the past year. For the distribution of bill payment amounts in the past year of the quantitative data mode with a small number of samples, it is necessary to use oversampling technology (such as SMOTE, ADASYN, etc.) to generate new synthetic samples for appropriate data balancing. Specifically, according to the supply chain management data The data balancing algorithm received from the database implements data balancing, that is, uses the SMOTE algorithm to obtain more sample data of the distribution of bill payment amounts in the past year of the quantitative data mode, and generates new synthetic samples according to the K nearest neighbor algorithm; and adds the generated samples to the original data set; after balancing, the data distribution is reanalyzed to ensure that the number of samples reaches a balanced state for subsequent calls), and multiple data in a certain category or multiple categories of data modes corresponding to the preprocessing completion quality index greater than or equal to the preset preprocessing completion quality threshold do not need to be balanced, and the following structured process processing can be directly performed; and various data modes that have completed the structured preprocessing operation are subjected to structured process processing, thereby generating a supply chain management knowledge graph.
[0043] Through the supply chain management knowledge graph generated above, the operation test of the knowledge graph can be carried out to obtain a test result chart, wherein the test result chart is a display form in which the operation result of the supply chain management knowledge graph constructed by the supply chain management database is integrated into a display form through the above detailed structured process processing, and this display form is the test result chart (the specific integration process is not described in detail in this example); and the data in the test result chart is a plurality of data after the above preprocessing, for example, the user selects the region, industry, industry, bill type, etc. to enter the key fields in the graph interface, so that the data in the test result chart is the total number of enterprises (using bills), the distribution of the years of establishment of enterprises, the distribution of registered capital, the distribution of the industry to which they belong, the distribution of the amount of bills paid in the past year, the distribution of the frequency of bills paid in the past year, the distribution of the number of bill circulations, and the distribution of the number of overdue bill payments; wherein the test result chart has the characteristic that the data classification corresponds to the data color one by one, that is, if there are 5 types of data, there will be 5 corresponding colors in the test result chart, so that the user can understand it intuitively; wherein the 5 types of data in the test result chart are arranged horizontally in the form of a table, so the area of the area to which each type of test label belongs below can be the area occupied by the table of the 5 types of data.
[0044] Specifically, the test result of the supply chain management knowledge graph is obtained by: The test query result data corresponding to each test user is specifically the test result chart quality data corresponding to each test user, including the color saturation of each test label in each test result chart, the area of each test label in each test result chart, and the average sharpness of the area of each test label in each test result chart.
[0045] It should be explained that users will obtain supply chain management data in the form of charts through the knowledge graph. Therefore, in order to clearly distinguish the types of data in the chart, the supply chain management database will divide the multiple types of data displayed in the chart into colors, that is, each type of data corresponds to a color, so the above-mentioned test label is a type of data; the color saturation corresponding to the test label can be obtained by using visualization tools (such as Matplotlib, Seaborn, etc.), and the average sharpness of the area to which the test label belongs is obtained through image processing technology, that is, it is calculated using methods such as image gradient and Laplace operator (the calculation process is not described in detail here). The calculated sharpness is the sharpness between the upper limit of the area to which a certain test label belongs and the lower limit of the area to which an adjacent test label belongs. If the area to which a certain test label belongs has no adjacent test label with an upper limit, then the average sharpness of the area to which the test label belongs is the average sharpness of its own lower limit. Similarly, if the area to which a certain test label belongs has no adjacent test label with a lower limit, then the average sharpness of the area to which the test label belongs is the average sharpness of its own upper limit. The area to which the test label belongs here is the line segment that the test result chart uses to divide the data, which is recorded as an area. Since the chart of the knowledge graph will describe the situation of each type of data in detail, the area of the area to which the test label belongs can be obtained directly from the chart.
[0046] The total area of each test result chart can be directly obtained from the supply chain management database; and based on the color saturation of each test label of each test result chart, the area of each test label of each test result chart, and the average sharpness of each test label of each test result chart, a comprehensive analysis is performed to obtain the test evaluation value corresponding to each test result chart. The obtained method is as follows: In the formula, The color saturation of the qth test label in the cth test result chart is an important attribute of color purity and intensity, which reflects the brightness and vitality of the color.
[0047] is the area of the region to which the qth type of test label belongs in the cth test result chart, indicating that the data with this type of test label occupies the entire area of the chart.
[0048] is the total area of the cth test result chart, indicating the overall area size of the chart.
[0049] The reference ratio of the area of the test labels preset in the graph generation management platform indicates the reasonable ratio of the area of a certain type of test labels to the area of the chart.
[0050] is the average sharpness of the area to which the qth test label of the cth test result chart belongs, which is an indicator of the image clarity or detail performance in a specific area.
[0051] The influence factors corresponding to the sharpness unit values preset in the atlas generation management platform can be directly obtained from the atlas generation management platform when used, and the corresponding relationship can be a preset mapping relationship. For example, the regional average sharpness and the influence factors corresponding to the sharpness unit values preset in the atlas generation management platform form a mapping set, and the real-time regional average sharpness mapping set is used to obtain the influence factors corresponding to the sharpness unit values. The mapping relationship can be one-to-one or many-to-one. In this example, the value range is [0, 1].
[0052] c is the number of each test result chart, , y is the total number of test result charts, q is the number of each test label, , u is the total category of the test label, e is a natural constant, is the test evaluation value corresponding to the cth test result chart. In the process of testing the knowledge graph in this example, the final chart obtained will directly affect whether the quality of the graph is good. That is, if the value of a certain type of color saturation deviates from the value of the overall color saturation, it will indicate that the color of this type of data is significantly different from the color of the rest of the data, further indicating that the quality of generating this type of data is poor, thereby affecting the quality of the entire knowledge graph. If the quality of the graph is poor, it will directly cause the sharpness of a certain type of data to become smaller, that is, there will be unclear situations, making it impossible to intuitively understand the area proportion of this type of data in the overall chart, that is, there will be a certain deviation between the value of the area proportion and the actual value, which is ultimately not conducive to the user's overall understanding of the results generated by the chart, so that wrong supply chain management decisions are made.
[0053] The test process data of each test user specifically includes the chart generation time of each test user, the number of incorrect input instructions of each test user in the test input period, and the number of input fields of each test user at the end time point of the test input period, and the corresponding test interference value of each test user is obtained through data processing.
[0054] The chart generation time of each of the above-mentioned test users, the number of incorrect input instructions of each test user in the test input period, and the number of input fields of each test user at the end time point of the test input period can all be directly extracted from the user application interface of the supply chain management database; the test input period represents the time period when the test user inputs key fields, and the specific duration will be determined according to the number of user input fields.
[0055] Among them, each test user corresponds to a test interference value, and the method of obtaining it is as follows: in, It is the chart generation time for the ath test user. In this example, since the test user will only get one chart after the test query of the knowledge graph, the above test result charts and the test users here have the same number; the chart generation time represents the overall time period from the execution of the chart generation instruction to the acquisition of the complete chart of the knowledge graph.
[0056] The impact factor corresponding to the unit value of the chart generation time preset in the graph generation management platform can be directly obtained from the graph generation management platform when used, and the corresponding relationship can be a preset mapping relationship. For example, the chart generation time and the impact factor corresponding to the unit value of the chart generation time preset in the graph generation management platform form a mapping set, and the real-time chart generation time mapping set is used to obtain the impact factor corresponding to the unit value of the chart generation time. The mapping relationship can be one-to-one or many-to-one. In this example, the value range is [0, 1].
[0057] It is the number of incorrect input instructions of the ath test user during the test input period, indicating that after the test user enters a certain field, the knowledge graph will prompt the incorrect input field instruction.
[0058] The impact factor corresponding to the unit value of the number of input error instructions is preset in the graph generation management platform. When used, the impact factor corresponding to the unit value of the number of input error instructions can be directly obtained from the graph generation management platform. The corresponding relationship can be a preset mapping relationship. For example, the number of input error instructions and the impact factor corresponding to the unit value of the number of input error instructions preset in the graph generation management platform form a mapping set, and the real-time input error instruction number mapping set is used to obtain the impact factor corresponding to the unit value of the number of input error instructions. The mapping relationship can be one-to-one or many-to-one. In this example, its value range is [0, 1].
[0059] It is the number of input fields of the ath test user at the end time point of the test input period, indicating the total number of input fields corresponding to the chart generation that the test user can finally perform. The field represents a keyword, which can be a single word or a keyword composed of multiple words.
[0060] The total number of input fields preset for the supply chain management database indicates the maximum number of fields that can be input by users as specified in the supply chain management database.
[0061] The reference ratio of input fields preset for the graph generation management platform indicates a reasonable ratio of the number of input fields to the total number of input fields.
[0062] e is a natural constant, a is the number of each test user, , m is the number of test users.
[0063] is the test interference value corresponding to the ath test user. In this example, since the test results of the knowledge graph are not only affected by the chart itself, but also by the test process of the test user corresponding to the chart, it will directly affect the quality of the chart. Therefore, if the user inputs incorrect instructions many times during the test input period, it means that when the user uses the knowledge graph, there will be certain error information in the fields he inputs, and these error information will still be recorded by the knowledge graph, which will interfere with the generation result of the chart. At the same time, if the user has a large number of input fields, the knowledge graph needs to use all the user's input fields in the process of generating the chart, and will also add the user's incorrect input information, which will ultimately make the chart generation time longer and the quality of the generated chart will also have certain deviations. Therefore, through the above description, it can be known that the test process of the test user needs to be analyzed so that when the supply chain management knowledge graph is fed back later, the existing problems can be found out in a timely and accurate manner.
[0064] Through the test evaluation values corresponding to each test result chart and the test interference values corresponding to each test user, the test results of the supply chain management knowledge graph are obtained through comprehensive analysis. The specific method is as follows: in, is the test evaluation value corresponding to the cth test result chart, which represents the value obtained by comprehensive analysis of color saturation, regional area and regional average sharpness.
[0065] The test evaluation value corresponding weight factor is preset for the graph generation management platform. When used, the test evaluation value corresponding weight factor can be directly obtained from the graph generation management platform. The corresponding relationship can be a preset mapping relationship. For example, color saturation and the test evaluation value corresponding weight factor preset in the graph generation management platform form a mapping set. The real-time color saturation mapping set is used to obtain the test evaluation value corresponding weight factor. The mapping relationship can be one-to-one or many-to-one. In this example, the value range is [0, 1].
[0066] is the test interference value corresponding to the a-th test user, which is the value obtained by comprehensive analysis of the chart generation time, the number of incorrect input instructions, and the number of input fields.
[0067] The weight factors corresponding to the test interference values are preset for the graph generation management platform. When used, the weight factors corresponding to the test interference values can be directly obtained from the graph generation management platform. The corresponding relationship can be a preset mapping relationship. For example, the graph generation time and the weight factors corresponding to the test interference values preset in the graph generation management platform form a mapping set. The real-time graph generation time mapping set is used to obtain the weight factors corresponding to the test interference values. The mapping relationship can be one-to-one or many-to-one. In this example, the value range is [0, 1].
[0068] For the test results of the supply chain management knowledge graph, in order to obtain an accurate and high-quality knowledge graph, it is necessary to test it, so the test process and the test results will affect its quality. That is, if there is a deviation in the test process (that is, the above-mentioned chart generation time is too long or the number of incorrect instructions is input is large or the number of input fields deviates from the reasonable number of fields), it will affect the quality of the generated chart, so the chart may be of poor quality (that is, the above-mentioned color saturation deviates from the mean or the area ratio deviation is large or the area sharpness is small), which will eventually lead to a relatively low test result of the knowledge graph and require further feedback.
[0069] e is a natural constant, c is the number of each test result chart, , y is the total number of test result charts, and a is the number of each test user. , m is the number of test users.
[0070] Furthermore, the test of the supply chain management knowledge graph is specifically that each test user in the supply chain management test scenario performs a graph query test through the supply chain management knowledge graph, and collects the corresponding test query result data of each test user and the test process data of each test user.
[0071] Among them, each test user in the above supply chain management test scenario is a part of the registered users who have been agreed in advance by the supply chain management database to use the knowledge graph. Therefore, the above test users receive the graph query test instructions sent by the supply chain management database and perform graph query tests through the supply chain management knowledge graph.
[0072] Specifically, the test of the supply chain management knowledge graph is fed back, and the specific feedback process is as follows: The test results of the supply chain management knowledge graph are compared with the test threshold preset by the graph generation management platform, where the test threshold represents the minimum value of the standard result obtained by testing the knowledge graph. If the test result of the supply chain management knowledge graph is greater than or equal to the preset test threshold, the supply chain management knowledge graph is visualized. Specifically, the supply chain management database sends the above-mentioned supply chain management knowledge graph to the interactive visualization tool, thereby helping users to intuitively understand the relevant situation of the supply chain, such as the graph display of the supply chain links, etc. The interactive visualization tool is designed with a dynamic update mechanism, thereby realizing automatic updating of real-time data to ensure that the graph reflects the latest supply chain status.
[0073] If the test result of the supply chain management knowledge graph is less than the preset test threshold, the test of the supply chain management knowledge graph is fed back through the feedback mechanism in the supply chain management database. Specifically, feedback information from users and the supply chain management database operation is collected to continuously optimize the knowledge graph (for example, if the feedback information is that the color saturation of the test result chart corresponding to the knowledge graph is significantly different, then the color saturation of the test result chart corresponding to the knowledge graph needs to be optimized so that the difference is within the allowable deviation range. The specific size of the above color saturation difference and the amount of color saturation optimization need to be combined with the specific feedback color saturation of the feedback information, which is not explained in detail here), so that the knowledge graph can not only meet the user's intuitive viewing needs, but also show high-quality data management.
[0074] In a specific embodiment, the present invention adopts advanced structured processing technology to generate a high-quality supply chain management knowledge graph, and combines it with the supply chain management test scenario to test the supply chain management knowledge graph, thereby obtaining a supply chain management knowledge graph that satisfies the users and can provide users with comprehensive, real-time and accurate information support, ensuring that the graph is effectively applied to the supply chain scenario to improve the information management level of the supply chain.
[0075] Reference Figure 2 As shown, the second aspect of the present invention provides a system for data structured processing and graph generation methods for supply chain management, including: a data preprocessing module, a knowledge graph generation module, a graph test feedback module and a graph generation management platform.
[0076] The graph generation management platform stores multiple preset values, multiple influencing factors and weight factors, which will not be described in detail here.
[0077] The data preprocessing module is connected to the knowledge graph generation module, the knowledge graph generation module is connected to the graph test feedback module, and the data preprocessing module, the knowledge graph generation module and the graph test feedback module are all connected to the graph generation management platform.
[0078] The data preprocessing module is used to collect supply chain related data of various data modes based on the established supply chain management database, and perform preprocessing operations, thereby determining the preprocessing completion quality indicators of various data modes.
[0079] The knowledge graph generation module is used to compare the preprocessing completion quality indicators of various data modalities with the preset preprocessing completion quality thresholds to obtain the comparison results, and perform structured process processing based on the comparison results, thereby generating a supply chain management knowledge graph.
[0080] The graph test feedback module is used to test the supply chain management knowledge graph through the supply chain management knowledge graph and in combination with the supply chain management test scenario, thereby obtaining the test results of the supply chain management knowledge graph and providing feedback on the test of the supply chain management knowledge graph.
[0081] In a specific embodiment, the present invention provides a data structured processing and graph generation method and system for supply chain management. First, detailed preprocessing operations are performed on the collected supply chain-related data, and the preprocessing completion quality indicators of various data modalities are determined, which can significantly improve the quality and consistency of multimodal data. Thus, a more comprehensive and accurate supply chain management knowledge graph is generated through structured process processing. The supply chain management knowledge graph is tested by combining the supply chain management test scenario to obtain the test results of the supply chain management knowledge graph. The generated supply chain management knowledge graph meets user needs through a feedback mechanism. At the same time, the high-quality supply chain graph provides accurate business insights and prediction capabilities, which ultimately enables users to accurately understand the overall picture of the relevant supply chain, providing a basis for making accurate and comprehensive supply chain management decisions in the future.
[0082] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.
Claims
1. A data structured processing and graph generation method for supply chain management, characterized in that: include: Based on the established supply chain management database, supply chain related data of various data modalities are collected and preprocessed, thereby determining the preprocessing completion quality indicators of various data modalities; Compare the preprocessing completion quality indicators of various data modalities with the preset preprocessing completion quality thresholds to obtain the comparison results, and perform structured process processing based on the comparison results to generate a supply chain management knowledge graph; Through the supply chain management knowledge graph, combined with the supply chain management test scenario, the supply chain management knowledge graph is tested, thereby obtaining the test results of the supply chain management knowledge graph and providing feedback on the test of the supply chain management knowledge graph.
2. The method for data structured processing and graph generation for supply chain management according to claim 1, characterized in that: The preprocessing operation specifically refers to the preprocessing operation module of the supply chain management database preprocessing the supply chain related data of various data modes, wherein the preprocessing operation module includes data cleaning, noise filtering and standardization.
3. The method for data structured processing and graph generation for supply chain management according to claim 2, characterized in that: The data cleaning specifically involves cleaning the supply chain-related data of various data modalities, wherein the supply chain-related data of various data modalities include various supply chain data points of various data modalities, and after cleaning the various supply chain data points of various data modalities, cleaning values corresponding to the various supply chain data points of various data modalities are generated; Compare the cleaning values corresponding to each supply chain data point of each data modality with the preset cleaning threshold. If the cleaning value corresponding to a supply chain data point of a certain data modality is less than the preset cleaning threshold, the supply chain data point of this data modality is recorded as an abnormal supply chain data point of this data modality, and provide data cleaning abnormality feedback; If the cleaning value corresponding to a supply chain data point of a certain type of data modality is greater than or equal to the preset cleaning threshold, then the supply chain data point of this type of data modality is recorded as the normal supply chain data point of this type of data modality. Thus, a number of supply chain data points corresponding to the cleaning values greater than or equal to the preset cleaning threshold are counted and recorded as the normal supply chain data points of each type of data modality.
4. The method for data structured processing and graph generation for supply chain management according to claim 3, characterized in that: The specific feedback process of data cleaning abnormality feedback is as follows: Based on the comparison results of cleaning values less than the preset cleaning threshold, a number of supply chain data points corresponding to cleaning values less than the preset cleaning threshold are counted and recorded as abnormal supply chain data points of various data modes; Perform difference processing on the preset cleaning threshold and the cleaning value corresponding to each abnormal supply chain data point of each data modality, obtain the cleaning deviation value corresponding to each abnormal supply chain data point of each data modality, and match it with the data cleaning adjustment plan corresponding to each preset cleaning deviation value interval, thereby obtaining the data cleaning adjustment plan corresponding to each abnormal supply chain data point of each data modality; Based on the data cleaning adjustment plan corresponding to each abnormal supply chain data point of each data modality, each abnormal supply chain data point of each data modality is re-cleaned according to the corresponding data cleaning adjustment plan to obtain the corresponding adjustment cleaning value of each abnormal supply chain data point of each data modality, and compare it with the preset cleaning threshold. If the corresponding adjustment cleaning value of a certain abnormal supply chain data point of a certain data modality is less than the preset cleaning threshold, the abnormal supply chain data point of this data modality is eliminated; If the adjusted cleaning value corresponding to an abnormal supply chain data point of a certain type of data modality is greater than or equal to the preset cleaning threshold, then the abnormal supply chain data point of this type of data modality is recorded as the normal adjusted supply chain data point of this type of data modality, thereby counting the normal adjusted supply chain data points of each type of data modality.
5. The method for data structured processing and graph generation for supply chain management according to claim 1, characterized in that: The specific determination process of determining the preprocessing completion quality index of each data modality is as follows: Each normal supply chain data point of each data modality and each normal adjusted supply chain data point of each data modality are subjected to noise filtering and standardized preprocessing operations in turn. Based on the completion process of noise filtering and standardized preprocessing operations, the preprocessing completion quality indicators of each data modality are comprehensively determined.
6. The data structured processing and graph generation method for supply chain management according to claim 1, characterized in that: Based on the comparison results, structured process processing is performed to generate a supply chain management knowledge graph. The specific process is as follows: Based on the comparison results, if the comparison results show that the preprocessing completion quality index of each data modality is greater than or equal to the preset preprocessing completion quality threshold, it is determined that each data modality is directly processed by the structured process and a supply chain management knowledge graph is generated; If the comparison results show that the preprocessing completion quality indicators of various data modalities are less than the preset preprocessing completion quality threshold, it is necessary to perform structured preprocessing operations on various data modalities, and perform structured process processing on various data modalities that have completed the structured preprocessing operations, thereby generating a supply chain management knowledge graph.
7. The method for data structured processing and graph generation for supply chain management according to claim 1, characterized in that: The test of the supply chain management knowledge graph is specifically that each test user in the supply chain management test scenario performs a graph query test through the supply chain management knowledge graph, and collects the corresponding test query result data of each test user and the test process data of each test user.
8. The method for data structure processing and graph generation for supply chain management according to claim 7, characterized in that: The test results of the supply chain management knowledge graph are obtained in the following specific process: The test query result data corresponding to each test user, specifically, the quality data of the test result chart corresponding to each test user, including the color saturation of each test label of each test result chart, the area of each test label of each test result chart, and the average sharpness of the area of each test label of each test result chart; Obtain the total area of each test result chart, and based on the color saturation of each test label of each test result chart, the area of each test label of each test result chart, and the average sharpness of each test label of each test result chart, comprehensively analyze and obtain the test evaluation value corresponding to each test result chart; The test process data of each test user specifically includes the chart generation time of each test user, the number of wrong input instructions of each test user in the test input period, and the number of input fields of each test user at the end time point of the test input period, and the test interference value corresponding to each test user is obtained through data processing; Through the test evaluation values corresponding to each test result chart and the test interference values corresponding to each test user, the test results of the supply chain management knowledge graph are obtained through comprehensive analysis. The specific method is as follows: in, is the test evaluation value corresponding to the c-th test result chart, The corresponding weight factors of the test evaluation values preset for the graph generation management platform, is the test interference value corresponding to the a-th test user, The corresponding weight factor of the test interference value preset for the graph generation management platform, is the test result of the supply chain management knowledge graph, e is a natural constant, c is the number of each test result chart, , y is the total number of test result charts, a is the number of each test user, , m is the number of test users.
9. The method for data structure processing and graph generation for supply chain management according to claim 1, characterized in that: The feedback on the test of the supply chain management knowledge graph is as follows: Compare the test result of the supply chain management knowledge graph with the preset test threshold. If the test result of the supply chain management knowledge graph is greater than or equal to the preset test threshold, visualize the supply chain management knowledge graph. If the test result of the supply chain management knowledge graph is less than the preset test threshold, the test of the supply chain management knowledge graph will be fed back through the feedback mechanism in the supply chain management database.
10. A system using the data structured processing and graph generation method for supply chain management as claimed in any one of claims 1 to 9, characterized in that: include: The data preprocessing module is used to collect supply chain related data of various data modes based on the established supply chain management database, and perform preprocessing operations, thereby determining the preprocessing completion quality indicators of various data modes; The knowledge graph generation module is used to compare the preprocessing completion quality indicators of various data modalities with the preset preprocessing completion quality thresholds to obtain the comparison results, and perform structured process processing based on the comparison results to generate a supply chain management knowledge graph; The graph test feedback module is used to test the supply chain management knowledge graph through the supply chain management knowledge graph and in combination with the supply chain management test scenario, thereby obtaining the test results of the supply chain management knowledge graph and providing feedback on the test of the supply chain management knowledge graph.
Citation Information
Patent Citations
Self-analysis radar rainfall estimating method based on big data
CN105717556A
Real-time acquisition and intelligent processing system based on industrial production and energy consumption data
CN114238477A
Supply chain knowledge graph analysis method based on big data
CN114254126A
Supply chain operation knowledge graph construction method and system
CN114610898A
Production intelligent feedback regulation and control system based on data mining
CN115630839A
Cited By
Intelligent supply chain management system and method based on artificial intelligence and big data
CN121581965A
