Information graph optimization system based on digital processing of industrial chain

Abstract

The application relates to the technical field of information graph optimization, and particularly discloses an information graph optimization system based on industrial chain digital processing, a logical architecture of the system comprises a data layer, a graph layer and an application layer, the application shortens the data processing link response time through a layered processing mechanism, realizes parallel processing of data quality checking and graph generation, a fuzzy data influence analysis unit provides dynamic monitoring capability, helps to identify potential risk points in a supply chain, an intelligent management unit executes targeted management and control based on quantitative analysis results, and the adaptive capability of the system to industrial chain fluctuations is enhanced.

Description

Technical Field

[0001] This invention relates to the field of information graph optimization technology, specifically an information graph optimization system based on digital processing of the industrial chain. Background Technology

[0002] In the process of digitalizing the industrial chain, traditional information mapping methods suffer from problems such as low data processing efficiency and unstable map quality.

[0003] For example, the patent with application number 2023117579542 involves an automatic method and system for constructing an industry chain map. The specific technical content is as follows: the data acquisition module is responsible for data acquisition, the data processing module is responsible for data preprocessing, the data recognition module is responsible for recognizing entities and entity relationships in the data, the map construction module is responsible for constructing and optimizing the industry chain map, the map query module is responsible for querying the industry chain map, the data visualization module is responsible for displaying the industry chain map, and the map update module is responsible for updating the industry chain map regularly.

[0004] However, the aforementioned patent content only uses a single data pool to store all types of data, resulting in low efficiency in filtering redundant data. At the same time, traditional data preprocessing methods usually only filter outliers by using fixed thresholds, which cannot distinguish the underlying causes of data fluctuations. Furthermore, data verification methods still perform rule checks on a single data type, and simply detecting missing fields or analyzing numerical ranges cannot provide a thorough verification. Therefore, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to solve the problems mentioned above by proposing an information graph optimization system based on digital processing of the industrial chain.

[0006] The objective of this invention can be achieved through the following technical solution: This invention proposes an information graph optimization system based on digital processing of the industrial chain. The logical architecture of the system includes a data layer, a graph layer, and an application layer. The system includes an optimization platform, and the optimization platform is connected to a data preprocessing unit, a data quality verification unit, a fuzzy data impact analysis unit, and an application intelligent management unit.

[0007] The optimization platform is configured to perform information analysis and optimization based on the digital processing of the industrial chain, and generate data preprocessing signals;

[0008] The data preprocessing unit is configured to preprocess the real-time collected operational data of each link in the industrial chain in response to the data preprocessing signal, in order to distinguish redundant data from valid data, and generate a data quality verification signal after completion.

[0009] The data quality verification unit is configured to perform quality verification on valid data in response to a data quality verification signal, and send the valid data that passes the verification to the optimization platform.

[0010] The optimized platform is further configured to drive the graph layer to extract entity relationships, fuse knowledge, and generate graphs based on valid data that has passed verification, in order to construct a knowledge graph of the industrial chain and generate fuzzy data influence analysis signals upon completion.

[0011] The fuzzy data impact analysis unit is configured to respond to the fuzzy data impact analysis signal, analyze the fuzzy impact factors of data in the industrial chain knowledge graph, and output the analysis results to the optimization platform.

[0012] The application intelligent management unit is configured to perform application management based on data from the graph layer in response to application management signals.

[0013] Furthermore, the data preprocessing unit preprocesses the real-time collected operational data of each link in the industrial chain, specifically including: according to preset rules, marking the quantitative data of each link as either natural or artificially fluctuating.

[0014] For quantified data of the natural floating type, its value is compared with the floating range set based on historical data; if the value of the quantified data is outside the floating range and appears only once, the data is marked as redundant data and temporarily stored; if the value of the quantified data is within the floating range, the data is marked as valid data.

[0015] Furthermore, the data quality verification unit performs quality verification on valid data, specifically including: identifying valid data as hard data or floating data;

[0016] The system retrieves the number of missing characters in the required fields of the hard data and the numerical deviation between the frequency of the collected values ​​and the actual frequency of the floating values ​​within the floating period. If the number of missing characters exceeds the missing character threshold or the numerical deviation exceeds the frequency deviation threshold, the data quality verification is deemed unqualified, and unqualified data is removed and data source tracing is performed. If the number of missing characters does not exceed the missing character threshold and the numerical deviation does not exceed the frequency deviation threshold, the data quality verification is deemed qualified.

[0017] Furthermore, the graph layer performs graph construction, specifically including: performing entity relationship extraction, extracting entities and relationships between entities from the data, and constructing the initial structure of the multi-relationship graph;

[0018] The system performs knowledge fusion, semantically vectorizes textual data, generates low-dimensional entity representations by combining node structural features, constructs an industry chain knowledge graph, performs graph generation, integrates structured, semi-structured, and unstructured data, and generates an industry chain knowledge graph containing nodes, edges, and relationship weights.

[0019] Furthermore, the fuzzy data impact analysis unit analyzes the fuzzy influencing factors of data in the industrial chain knowledge graph, specifically including: transforming fuzzy influencing factors into observable data, which at least includes supply and demand fluctuation data and logistics delay data;

[0020] Each observable data point is divided into positive or negative indicators. The values ​​of each observable data point are collected to form a data state curve. Based on the comparison between the actual values ​​at each time point and the set values, the data state curve is color-coded. Based on the color distribution of all data state curves in the industry chain knowledge graph, the graph influence coefficient is calculated. If the graph influence coefficient exceeds the set ratio threshold, the fuzzy data is judged to have a high influence, and a data anomaly signal is generated. If the graph influence coefficient does not exceed the set ratio threshold, the fuzzy data is judged to have a low influence, and a data normal signal is generated.

[0021] Furthermore, the fuzzy data impact analysis unit is further configured to: calculate the color alternation frequency for the data state curve of a single type of data;

[0022] If the color alternation frequency exceeds the set frequency threshold, the fuzzy data is deemed to have a continuous impact, and an inefficient data control signal is generated. If the color alternation frequency does not exceed the set frequency threshold, the fuzzy data is deemed to have an occasional impact, and an efficient data control signal is generated.

[0023] Furthermore, the supply and demand fluctuation data includes supply-side fluctuation data and demand-side fluctuation data. The supply-side fluctuation data includes at least one of the following: core supplier capacity change rate, raw material price volatility, and inventory turnover deviation. The demand-side fluctuation data includes at least one of the following: year-on-year growth rate of terminal sales, order cancellation rate, and market demand forecast deviation rate.

[0024] Furthermore, logistics delay data includes transportation delay data and warehousing delay data. Transportation delay data includes at least one of the following: trunk line transportation time deviation, branch line delivery delay rate, and transportation interruption frequency. Warehousing delay data includes at least one of the following: inbound sorting time, outbound preparation time, and inventory turnover stagnation period.

[0025] Furthermore, after receiving a data anomaly signal, the optimization platform is further configured to manage the corresponding type of data, including controlling the fluctuation trend when the data value is in a fluctuating state and controlling the specific value when the data value is in a stable state.

[0026] Furthermore, upon receiving a signal of inefficient data control, the optimization platform is further configured to rectify the fuzzy data control process for the corresponding type of data, including data float control and data detection timing control.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. This invention effectively reduces the interference of redundant data on knowledge graph construction, improves data cleaning efficiency, shortens the response time of data processing links through a layered processing mechanism, realizes parallel processing of data quality verification and graph generation, provides dynamic monitoring capabilities through the fuzzy data impact analysis unit, helps identify potential risk points in the supply chain, and applies intelligent management units to perform targeted control based on quantitative analysis results, thereby enhancing the system's adaptability to fluctuations in the industrial chain.

[0029] 2. The present invention also effectively distinguishes between normal fluctuations and real anomalies in the industrial chain data through a data preprocessing unit, reducing the interference of redundant data on subsequent processing. For example, in the inventory management scenario, reasonable inventory fluctuations caused by seasonal demand growth will not be filtered out incorrectly, while single inventory anomalies caused by transportation failures will be identified in a timely manner. As a result, the accuracy of data cleaning is improved, providing higher quality data input for map construction.

[0030] 3. This invention also uses a data quality verification unit to simultaneously address two types of problems: insufficient integrity of hard data and abnormal fluctuations in floating data. This prevents low-quality data from entering the graph construction stage, thereby reducing the error rate of subsequent knowledge fusion and graph generation. For example, in a supply chain scenario, missing required fields may lead to incomplete supplier node information, while abnormal fluctuations in logistics delay data may distort the relationship between warehousing nodes. The dual verification of this solution can effectively block the propagation of such errors.

[0031] 4. This invention also effectively identifies potential risks caused by fuzzy data in the supply chain knowledge graph through a fuzzy data impact analysis unit. For example, when logistics delay data frequently triggers color mark changes in a short period of time, the system can quickly locate abnormal nodes in the transportation process. At the same time, by distinguishing between positive and negative indicators, the system can accurately determine the direction of the impact of data fluctuations on supply chain optimization. For example, when demand-side fluctuation data is determined to be a negative indicator and the impact coefficient exceeds the standard, the system can adjust the parameters of the demand forecasting model accordingly to improve the accuracy of knowledge graph construction and the reliability of decision-making. Attached Figure Description

[0032] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0033] Figure 1 This is a schematic diagram of the system logic architecture of the present invention;

[0034] Figure 2 This is a schematic diagram of the internal unit composition of the system of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0037] In existing technologies, the digital processing of the industrial chain involves the collection and integration of multi-source heterogeneous data. Traditional methods usually use a single data processing module for centralized management. Due to the lack of a layered architecture design, there is a problem of excessive coupling in the data preprocessing, quality verification and knowledge graph construction stages, resulting in low efficiency in filtering redundant data and difficulty in dynamically tracking the impact of ambiguous factors on the knowledge graph. For example, in the supply chain management scenario, real-time collected production capacity data and logistics data are stored together, and the identification of abnormal data is delayed, affecting the accuracy of supply and demand forecasting models.

[0038] To address these issues, researchers discovered that traditional systems could not effectively distinguish between naturally fluctuating data and anomalous data, and that the quality verification process did not consider data type differences. After analysis, they realized that a hierarchical processing mechanism must be established, with independent functional units handling data cleaning, quality verification, and impact analysis separately.

[0039] Further research revealed that the impact of fuzzy data on knowledge graphs has dynamic propagation characteristics, requiring the establishment of a closed-loop feedback mechanism. Based on this, it is proposed to divide the system into a data layer, a graph layer, and an application layer, and design dedicated units to achieve functional decoupling of each layer.

[0040] Please see Figures 1-2 As shown, the system logic architecture of the information graph optimization system based on digital processing of the industrial chain consists of a data layer, a graph layer, and an application layer.

[0041] The system includes an optimization platform, which is connected to a data preprocessing unit, a data quality verification unit, a fuzzy data impact analysis unit, and an application intelligent management unit. The data preprocessing unit and the data quality verification unit are execution units of the data layer in the logical architecture, the fuzzy data impact analysis unit is the execution unit of the graph layer in the logical architecture, and the application intelligent management unit is the execution unit of the application layer in the logical architecture.

[0042] Compared with existing technologies, traditional systems use a single data pool to store all types of data, resulting in low efficiency in filtering redundant data. In contrast, this invention decouples data preprocessing, quality verification, and graph construction through a layered architecture. The data preprocessing unit and the quality verification unit belong to the data layer and are specifically responsible for cleaning and verifying the raw data. The fuzzy data impact analysis unit, as a graph layer component, can dynamically track the impact of data fluctuations on the knowledge graph. Furthermore, existing technologies lack a quantitative analysis module for fuzzy factors, while this invention achieves dynamic impact assessment through state curve color marking and impact coefficient calculation.

[0043] This invention effectively reduces the interference of redundant data on knowledge graph construction, improves data cleaning efficiency, shortens the response time of the data processing link through a layered processing mechanism, realizes parallel processing of data quality verification and graph generation, provides dynamic monitoring capabilities through the fuzzy data impact analysis unit to help identify potential risk points in the supply chain, and applies intelligent management unit to perform targeted control based on quantitative analysis results, thereby enhancing the system's adaptability to fluctuations in the industrial chain.

[0044] When the system is running, it adopts a "general protocol + customized driver" mode, covering the mainstream data acquisition protocols in the industry chain, and supports dynamic expansion;

[0045] IoT device protocols: Built-in industrial-grade protocol parsers such as MQTT, CoAP, Modbus-TCP, and OPC UA are used to convert binary stream data (such as temperature, production capacity, and equipment status) from sensors and production line equipment into structured data in JSON format through protocol conversion.

[0046] ERP system protocols: Supports interfaces such as JDBC / ODBC, REST API, and SOAP API, and connects to mainstream ERP systems such as SAP, Yonyou, and Kingdee. It collects structured information such as purchase orders, inventory ledgers, and financial data by combining timed incremental synchronization (e.g., every 5 minutes) with real-time triggered synchronization (e.g., when orders change).

[0047] Public data source protocols: For unstructured data such as web pages, industry reports, and policy documents, it integrates HTTP / HTTPS crawler protocols, FTP file transfer protocols, and API interfaces (such as the National Bureau of Statistics API and industry database API) to achieve batch crawling and incremental updates of data in formats such as text, PDF, and Excel.

[0048] Customized Protocol Extension: Provides a protocol development SDK, enabling enterprises to write parsing drivers based on their own equipment's (such as customized production machine tools) private protocols, and connect to the gateway in a plug-in manner, reducing system adaptation costs.

[0049] The optimization platform performs information analysis and optimization based on the digital processing of the industrial chain, generates data preprocessing signals, and sends them to the data preprocessing unit.

[0050] After receiving the data preprocessing signal, the data preprocessing unit performs data preprocessing on the data collected by the generation chain.

[0051] Real-time collection of operational data from the supply chain, production, and sales processes. This operational data is represented by quantitative data generated by the operation of each link in the industrial chain, such as the supply volume and speed of the supply chain, the output fluctuations in the production process, and the supply and demand fluctuations in the sales process.

[0052] Set preset rules, obtain the fluctuation range of quantitative data of corresponding types for each link based on the historical operation process, analyze the quantitative data of corresponding types for each link, if the data type is natural fluctuation, that is, the influence of human factors is low, such as the actual deviation of the set speed of the supply chain, the actual deviation of the set speed of the production line conveyor belt, etc., this type is marked as natural fluctuation type; otherwise, if the data type is human fluctuation, that is, the influence of human factors is high, such as the supply departure time of the supply chain, the output fluctuation of the production line, this type is marked as human fluctuation type.

[0053] The numerical comparison of the naturally floating type of quantitative data is performed according to the preset rules. If the quantitative data does not fall within the preset rules during the numerical fluctuation stage and appears alone, this type of data is marked as redundant data, and the redundant data is marked as abnormal and temporarily stored.

[0054] If the preset rules apply, the corresponding data type will be marked as valid data;

[0055] It should be explained that the preset rules are used to filter the validity of the data, rather than to perform operational checks on the process in which the data is located.

[0056] After the data preprocessing unit completes the division of redundant data and valid data, it generates a data quality verification signal and sends it to the data quality verification unit.

[0057] Compared with existing technologies, traditional data preprocessing methods usually only filter out outliers by using fixed thresholds, and cannot distinguish the underlying causes of data fluctuations. For example, existing technologies may directly remove all data that exceeds the preset value, causing normal market fluctuations to be misjudged as abnormal. In contrast, this invention uses type classification and dynamic comparison with historical fluctuation ranges to retain reasonable fluctuation data and accurately identify occasional anomalies, thus avoiding the loss of effective information.

[0058] This invention effectively distinguishes between normal fluctuations and real anomalies in supply chain data, reducing the interference of redundant data on subsequent processing. For example, in inventory management scenarios, reasonable inventory fluctuations caused by seasonal demand growth will not be filtered out incorrectly, while single inventory anomalies caused by transportation failures will be identified in a timely manner. As a result, the accuracy of data cleaning is improved, providing higher quality data input for graph construction.

[0059] After receiving the data quality verification signal, the data quality verification unit performs valid data quality verification.

[0060] The valid data is classified into hard data and floating data. Hard data refers to parameters such as equipment specifications or enterprise codes in each stage, while floating data refers to parameters generated by the operation of equipment in each stage (such as temperature and speed).

[0061] The missing data character count of mandatory fields in the digital processing of the industrial chain is obtained. For example, the unified social credit code of an enterprise is an essential information processing parameter for the identity of an enterprise in the digital processing of the industrial chain.

[0062] When processing digital data in the industrial chain, the deviation between the frequency of data collected during the floating period and the actual frequency of data fluctuation is obtained. It needs to be explained that the frequency deviation indicates whether there are missing data collections or excessively high collection frequencies in the same area, which leads to a decrease in the accuracy of the data reflected in each link.

[0063] When processing the digital transformation of the supply chain, the numerical deviations between the missing character count and the actual numerical fluctuation frequency of the collected values ​​for the mandatory fields of hard data and floating data within the floating period are compared with the missing character count threshold and the frequency deviation threshold, respectively.

[0064] If the number of missing characters in the mandatory fields of hard data during the digital processing of the industrial chain exceeds the missing character threshold, or if the deviation between the frequency of the collected values ​​and the actual frequency of the floating data exceeds the frequency deviation threshold, it is inferred that the data quality verification during the digital processing of the industrial chain is unqualified. The currently valid data will be removed as unqualified data, and the collection process will be traced back. Furthermore, the values ​​will be extrapolated based on the data at adjacent times. It is only necessary to extrapolate whether the values ​​are missing and whether they are within the preset rules, without performing specific value extrapolation, so as to avoid value deviation.

[0065] If the number of missing characters in the mandatory fields of hard data during the digital processing of the industrial chain does not exceed the missing character threshold, and the numerical deviation between the collected value fluctuation frequency and the actual value fluctuation frequency during the floating data fluctuation period does not exceed the frequency deviation threshold, then it is inferred that the data quality verification during the digital processing of the industrial chain is qualified, and the current valid data is sent to the optimization platform.

[0066] Compared to existing technologies, traditional data verification methods typically only perform rule checks on a single data type, such as detecting missing fields or analyzing numerical ranges. In contrast, this invention distinguishes between hard data and floating data, designs verification indicators for each, and achieves multi-dimensional quality assessment. For example, existing technologies may ignore the periodic characteristics of floating data, leading to misjudging normal fluctuations as abnormal data. This invention, however, introduces frequency deviation calculations within the floating period, enabling more accurate identification of true anomalies.

[0067] This simultaneously addresses two types of problems: insufficient integrity of hard data and abnormal fluctuations in floating data. It prevents low-quality data from entering the graph construction stage, thereby reducing the error rate of subsequent knowledge fusion and graph generation. For example, in a supply chain scenario, missing required fields may lead to incomplete supplier node information, while abnormal fluctuations in logistics delay data may distort the relationship between warehouse nodes. The dual verification of this invention can effectively block the propagation of such errors.

[0068] After the optimization platform receives the valid data that has completed data layer processing, the operation logic of the graph layer is as follows:

[0069] Entity Relationship Extraction: Extract entities such as enterprises, equipment, and processes from standardized datasets, as well as the relationships between them, such as supply and demand relationships and cooperation relationships, to construct the initial structure of a multi-relationship graph;

[0070] Knowledge fusion: Semantic vectorization is performed on textual data, and word frequency vectors and co-occurrence matrices are mapped to the corresponding nodes of the multi-relationship graph. At the same time, combined with the structural features of the nodes, a low-dimensional representation of entities that integrates text and structural features is generated, thereby constructing an industry chain knowledge graph.

[0071] Graph generation: Based on the industry chain ontology, structured data, semi-structured text and unstructured data are integrated to generate an industry chain knowledge graph containing nodes, edges and relationship weights;

[0072] Compared with existing technologies, traditional knowledge graph construction methods usually only process structured data and have difficulty integrating semi-structured and unstructured data sources, resulting in incomplete coverage of entity relationships in the industrial chain. This invention, however, performs entity extraction, semantic fusion and heterogeneous data integration in stages, so that textual terms in procurement contracts, semi-structured waybills in logistics systems and structured records in production databases can all be transformed into graph elements, thereby solving the semantic gap problem when fusing multi-source data.

[0073] This enables unified graph modeling of all data in the industrial chain, allowing cross-system data such as supplier capacity fluctuations and logistics delays to be correlated and analyzed, providing complete data support for subsequent identification of fuzzy influencing factors. At the same time, it reduces the computational complexity of unstructured data processing and improves the efficiency of building large-scale industrial chain knowledge graphs by generating low-dimensional entity representations.

[0074] After the map construction is completed based on the map layer, a fuzzy data influence analysis signal is generated and sent to the fuzzy data influence analysis unit.

[0075] After receiving the fuzzy data impact analysis signal, the fuzzy data impact analysis unit processes and analyzes the fuzzy impact factors of the constituent data within the knowledge graph.

[0076] The fuzzy factors are transformed into observable and measurable data. The fuzzy influencing factors in this application take supply and demand fluctuations and logistics delays as the main data. The data type can be determined according to the usage scenario when the system is actually running.

[0077] Supply and demand fluctuations can be divided into two sub-dimensions: supply-side fluctuations and demand-side fluctuations.

[0078] Based on the sub-dimensions, observable data for the corresponding sub-dimensions are obtained. That is, based on supply-side fluctuations and demand-side fluctuations, corresponding supply-side fluctuation data and demand-side fluctuation data are collected. Supply-side fluctuation data include: core supplier capacity change rate, raw material price volatility, and inventory turnover deviation; demand-side fluctuation data include: year-on-year growth rate of terminal sales, order cancellation rate, and market demand forecast deviation rate.

[0079] Logistics delays are divided into two sub-dimensions: transportation delays and warehousing delays.

[0080] Obtain observable data for the corresponding sub-dimension based on the sub-dimension, that is, collect corresponding transportation delay data and warehousing delay data based on transportation delay and warehousing delay. Transportation delay data includes: trunk line transportation time deviation, branch line delivery delay rate, and transportation interruption frequency; warehousing delay data includes: inbound sorting time, outbound preparation time, and inventory turnover stagnation period.

[0081] The observable data corresponding to each sub-dimension are divided into positive and negative indicators. When the data increases, the operation of the knowledge graph corresponding to the industrial chain is better. For example, the higher the inventory turnover rate, the better the operation efficiency of the industrial chain. On the other hand, when the data increases, the operation of the knowledge graph corresponding to the industrial chain is worse. For example, when the capacity change rate of core suppliers increases, the operation of the corresponding industrial chain is worse.

[0082] Numerical data is collected based on observable data within the knowledge graph to form a data state curve;

[0083] The actual values ​​of positive and negative indicators at each time point are obtained from the data status curve. The positive and negative deviations are obtained by comparing the actual values ​​with the set values ​​for the current stage. When the positive deviation occurs above the set values, the data status curve corresponding to the current data of the knowledge graph is marked in green. When the positive deviation occurs below the set values, the data status curve corresponding to the current data of the knowledge graph is marked in red.

[0084] When the value exceeds the set value and a reverse deviation occurs, the data status curve corresponding to the current data of the knowledge graph is marked in red; when the value falls below the set value and a reverse deviation occurs, the data status curve corresponding to the current data of the knowledge graph is marked in green.

[0085] After setting the data status curve color through 3D rendering, the corresponding curve color distribution is performed on the entire knowledge graph data. Based on the cumulative duration of the red curve corresponding to the curve process time and the cumulative duration of the green curve corresponding to the curve process time in the knowledge graph, the duration ratio is obtained and set as the graph influence coefficient.

[0086] If the duration ratio exceeds the set ratio threshold, it is inferred that the influence of fuzzy data is high, generating a data anomaly signal and sending it to the optimization platform along with the corresponding type of data. The optimization platform manages this type of data, that is, when the data value is in a floating state, the floating trend is controlled, and when it is in a stable state, the specific value is controlled.

[0087] Conversely, if the duration ratio does not exceed the set ratio threshold, it is inferred that the influence of fuzzy data is low, a data normal signal is generated and sent to the optimization platform along with the corresponding type of data. The ratio threshold can be manually set by combining the data error rate with the duration of the actual curve process.

[0088] Based on the curve color comparison of a single type of data, the frequency of color alternation between the red and green curves in the state curve of a single data is obtained. If the frequency of color alternation exceeds the set frequency threshold, it is inferred that the influence of fuzzy data is continuous, and an inefficient data control signal is generated and sent to the optimization platform. After optimization, the fuzzy data control of this type of data is rectified, such as data float control and data detection time control.

[0089] If the color alternation frequency does not exceed the set frequency threshold, it is inferred that the fuzzy data affects occasional events, and a data control efficiency signal is generated and sent to the optimization platform.

[0090] Compared with existing technologies, traditional methods usually only perform static threshold judgments on a single data point, which cannot reflect the cumulative effect of ambiguity factors over time. However, this invention achieves continuous monitoring of the impact of ambiguity through dynamic data state curves and color marking mechanisms, and performs global evaluation by combining the spectral influence coefficient, thus upgrading anomaly judgment from isolated event detection to systemic risk identification.

[0091] This effectively identifies potential risks caused by fuzzy data in the supply chain knowledge graph. For example, when logistics delay data frequently triggers color-coded changes within a short period, the system can quickly locate abnormal nodes in the transportation process. Furthermore, by distinguishing between positive and negative indicators, it can accurately determine the direction of data fluctuations' impact on supply chain optimization. For instance, when demand-side fluctuation data is identified as a negative indicator and its impact coefficient exceeds the standard, the system can adjust the parameters of the demand forecasting model accordingly, thereby improving the accuracy of knowledge graph construction and the reliability of decision-making.

[0092] The application layer performs application management based on the data from the graph layer, generates application management signals, and sends them to the application intelligent management unit.

[0093] After receiving the data, the intelligent management unit models the nodes in the industry chain and assigns rendering priorities based on business importance.

[0094] Core layer nodes: These include leading enterprises in the industry chain, core products, and key equipment, accounting for approximately 10%-15% of the total number of nodes. They are given the highest rendering priority and are always visible.

[0095] Related layer nodes: Direct upstream and downstream partners of core nodes, accounting for approximately 30%-40% of the total number of nodes. They are visible by default and can be manually hidden by the user.

[0096] Extended layer nodes: Indirectly related SMEs and auxiliary products, accounting for approximately 45%-60% of the total number of nodes. They are hidden by default and are displayed by "hover expansion" or "layer scaling".

[0097] During the initial load, only the core layer and related layer nodes (approximately 200-300) are rendered. After loading is complete, the extension layer nodes are loaded asynchronously in the background to avoid page lag caused by loading all at once.

[0098] Thresholds, preset values, preset ranges, etc. are set for result comparison and analysis to determine whether they are good or bad. The size of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience. They can also be adjusted appropriately based on seasonal or common-sense influences.

[0099] Furthermore, the settings for weighting ratios, influence factors, etc., are based on the magnitude of each parameter's influence on the results. The specific values ​​are allocated to ultimately reflect the impact on the results. The settings for input and storage are also determined by a combination of large-scale model analysis of sample data and human experience. Appropriate adjustments can also be made based on seasonal or rational influence conditions.

[0100] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An information graph optimization system based on digital processing of the industrial chain, characterized in that, The logical architecture of the system includes a data layer, a graph layer, and an application layer. The system includes an optimization platform, and the optimization platform is communicatively connected to a data preprocessing unit, a data quality verification unit, a fuzzy data impact analysis unit, and an application intelligent management unit. The optimization platform is configured to perform information analysis and optimization based on the digital processing of the industrial chain, and to generate data preprocessing signals; The data preprocessing unit is configured to respond to the data preprocessing signal, preprocess the real-time collected operational data of each link in the industrial chain, distinguish between redundant data and valid data, and generate a data quality verification signal after completion. The data quality verification unit is configured to perform quality verification on the valid data in response to the data quality verification signal, and send the valid data that passes the verification to the optimization platform. The optimization platform is further configured to drive the graph layer to extract entity relationships, fuse knowledge, and generate graphs based on the verified valid data, so as to construct an industrial chain knowledge graph and generate fuzzy data influence analysis signals after completion. The fuzzy data impact analysis unit is configured to respond to the fuzzy data impact analysis signal, analyze the fuzzy impact factors of the data in the industrial chain knowledge graph, and output the analysis results to the optimization platform. The application intelligent management unit is configured to perform application management based on the data of the graph layer in response to application management signals; The data preprocessing unit preprocesses the real-time collected operational data from each link of the industrial chain, specifically including: According to preset rules, the quantitative data of each stage are marked as either natural fluctuation type or artificial fluctuation type; For the quantitative data of the aforementioned natural float type, its value is compared with the float range set based on historical data; If the value of the quantified data is outside the floating range and appears only once, then the data is marked as redundant data and temporarily stored. If the value of the quantified data is within the fluctuation range, then the data is marked as valid data; The data quality verification unit performs quality verification on the valid data, specifically including: The valid data is identified as either hard data or floating data; Obtain the number of missing data characters in the required fields of the hard data, and obtain the numerical deviation between the frequency of the collected values ​​and the actual frequency of the floating values ​​within the floating period of the floating data. If the number of missing data characters exceeds the missing character threshold, or the numerical deviation exceeds the frequency deviation threshold, the data quality verification is deemed unqualified, and unqualified data removal and data source tracing calculation are performed. If the number of missing data characters does not exceed the missing character threshold and the numerical deviation does not exceed the frequency deviation threshold, then the data quality verification is deemed qualified. The graph layer performs graph construction, specifically including: Perform entity relationship extraction, extract entities and relationships between entities from the data, and construct the initial structure of the multi-relationship graph; Perform knowledge fusion, semantically vectorize textual data, generate low-dimensional entity representations by combining node structural features, and construct an industry chain knowledge graph. Perform graph generation, integrate structured, semi-structured and unstructured data to generate an industry chain knowledge graph containing nodes, edges and relationship weights; The fuzzy data impact analysis unit analyzes the fuzzy influencing factors of the data in the industrial chain knowledge graph, specifically including: The ambiguous influencing factors are transformed into observable data, which includes at least supply and demand fluctuation data and logistics delay data; Each observable data point is categorized into positive or negative indicators. Collect values ​​of each observable data point to form a data state curve, and color-code the data state curve based on the comparison results between the actual values ​​at each time point and the set values. Calculate the graph influence coefficient based on the color distribution of all data state curves in the aforementioned industry chain knowledge graph; If the influence coefficient of the map exceeds the set ratio threshold, it is determined that the influence of fuzzy data is high, and a data anomaly signal is generated. If the influence coefficient of the spectral data does not exceed the set ratio threshold, the influence of the fuzzy data is determined to be low, and a normal data signal is generated. The fuzzy data impact analysis unit is further configured as follows: For the data state curve of a single type of data, calculate its color alternation frequency; If the color alternation frequency exceeds the set frequency threshold, it is determined that the fuzzy data affects the continuity, and an inefficient data control signal is generated. If the color alternation frequency does not exceed the set frequency threshold, it is determined that the fuzzy data influence is occasional, and a data control efficiency signal is generated. The supply and demand fluctuation data includes supply-side fluctuation data and demand-side fluctuation data. The supply-side fluctuation data includes at least one of the following: core supplier capacity change rate, raw material price volatility, and inventory turnover deviation. The demand-side fluctuation data includes at least one of the following: year-on-year growth rate of terminal sales, order cancellation rate, and market demand forecast deviation rate. The logistics delay data includes transportation delay data and warehousing delay data. The transportation delay data includes at least one of trunk line transportation time deviation, branch line delivery delay rate, and transportation interruption frequency. The warehousing delay data includes at least one of inbound sorting time, outbound preparation time, and inventory turnover stagnation period.

2. The information graph optimization system based on digital processing of the industrial chain according to claim 1, characterized in that, After receiving the data anomaly signal, the optimization platform is further configured to manage the corresponding type of data, including controlling the fluctuation trend when the data value is in a fluctuating state and controlling the specific value when the data value is in a stable state.

3. The information graph optimization system based on digital processing of the industrial chain according to claim 1, characterized in that, Upon receiving a signal indicating inefficient data control, the optimization platform is further configured to rectify the fuzzy data control process for the corresponding type of data, including data float control and data detection timing control.

Images