Knowledge graph-based semantic association and logic rule reasoning method

Abstract

The invention relates to the technical field of knowledge maps, and discloses a semantic association and logic rule inference method based on a knowledge map, and the method comprises the steps: obtaining a cross-functional department first data set, and constructing a cross-department knowledge map; extracting index nodes from the cross-department knowledge graph, identifying semantic conflict indexes, generating an index semantic conflict list, and constructing the cross-department knowledge graph after semantic mapping based on the index semantic conflict list; obtaining a second data set of a cross-functional department, and obtaining a fused cross-department knowledge graph and a semantic association degree matrix based on the second data set and the cross-department knowledge graph after semantic mapping; dynamically adjusting the weight of each node and edge of the fused cross-department knowledge graph according to the semantic association degree matrix, and generating a cross-department collaborative knowledge graph; generating a cross-department collaborative decision suggestion based on the cross-department collaborative knowledge graph; according to the invention, the efficiency and quality of cross-department collaborative decision-making are significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and more specifically, to a reasoning method based on semantic associations and logical rules of knowledge graphs. Background Art

[0002] In recent years, with the rapid development of industrial Internet and artificial intelligence, intelligent decision support systems based on knowledge graphs have been widely used in the industrial field. Especially in the field of smart factory manufacturing, cross-departmental collaborative optimization and decision-making face huge challenges. Some existing research works focus on the construction of knowledge graphs, trying to solve the problem of cross-departmental knowledge integration and application.

[0003] The Chinese patent application with publication number CN114610898A proposes a method and system for constructing a supply chain operation knowledge graph. The method first obtains data in the supply chain operation process, extracts and integrates knowledge, and constructs a supply chain operation knowledge graph. Based on the knowledge graph, knowledge semantic retrieval and knowledge precision push services are implemented. The method constructs a knowledge graph that integrates multiple data sources, supports fast query and semantic retrieval of data in different fields, and to a certain extent meets the user's needs for cross-stage knowledge association retrieval. However, the method does not involve how to solve the semantic conflict problem in the process of cross-departmental knowledge fusion, which may lead to deviations in the reasoning and decision-making process. In addition, the method lacks a dynamic weight adjustment mechanism and is difficult to adapt to complex and changeable cross-departmental collaborative optimization scenarios.

[0004] A Chinese patent with publication number CN118069856A discloses a knowledge graph construction method and application method. The method obtains multi-source heterogeneous data in the target field, and constructs a knowledge graph containing entities and relationships through steps such as data cleaning, integration, and semantic recognition. The method introduces a semantic recognition model, which can extract entities and relationships from unstructured data and expand the information source of the knowledge graph. However, the method mainly focuses on the construction of knowledge graphs in a single field, and lacks semantic mapping and conflict resolution mechanisms for cross-departmental knowledge. In practical applications, there may be semantic differences and conflicts in the knowledge graphs of different departments, which affects the accuracy of collaborative reasoning.

[0005] In summary, the existing knowledge graph construction methods lack cross-departmental semantic mapping and alignment mechanisms, making it difficult to effectively solve semantic conflicts between different departments, affecting the efficiency and accuracy of cross-departmental collaborative decision-making; the reasoning and decision-making process does not fully consider the semantic correlation between different departments, and fails to achieve cross-departmental collaborative optimization. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a reasoning method based on semantic associations and logical rules of knowledge graphs.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A reasoning method based on semantic associations and logical rules of knowledge graphs, including:

[0009] Obtain a first data set across functional departments, construct a structured data table based on the first data set, and construct a cross-departmental knowledge graph based on the structured data table;

[0010] Extract indicator nodes from the cross-departmental knowledge graph, identify semantic conflict indicators, and generate an indicator semantic conflict list; construct a semantically mapped cross-departmental knowledge graph based on the indicator semantic conflict list; obtain a second data set across functional departments, and obtain a fused cross-departmental knowledge graph and a semantic relevance matrix based on the second data set and the semantically mapped cross-departmental knowledge graph; extract the weights of each node and edge of the fused cross-departmental knowledge graph, and dynamically adjust the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix to generate a cross-departmental collaborative knowledge graph;

[0011] Based on the cross-departmental collaborative knowledge graph, cross-departmental collaborative decision-making recommendations are generated.

[0012] Furthermore, the first data set includes indicator definition documents, business process description documents and historical decision records of cross-functional departments; the cross-functional departments include production departments, quality management departments and supply chain departments;

[0013] The step of constructing a structured data table according to the first data set includes:

[0014] Extracting the indicator name A, the first relationship type and the indicator name B from the indicator definition document to form an indicator information triple; wherein the indicator name A is the subject of the indicator information triple, the first relationship type is the predicate of the indicator information triple, and the indicator name B is the object of the indicator information triple;

[0015] Extracting process node A', the second relationship type and process node B' from the business process description document to form a process triple; wherein process node A' is the subject of the process triple, the second relationship type is the predicate of the process triple, and process node B' is the object of the process triple;

[0016] Extracting decision events, third relation types, decision problems or solutions from historical decision records to form decision information triples; wherein the decision events are the subjects of the decision information triples, the third relation types are the predicates of the decision information triples, and the decision problems or solutions are the objects of the decision information triples;

[0017] Based on the indicator information triples, process triples and decision information triples, a structured data table is constructed. The structured data table includes triple subjects, triple predicates and triple objects.

[0018] Furthermore, the construction of a cross-departmental knowledge graph based on a structured data table includes:

[0019] Mapping triple subjects and triple objects into nodes in a cross-departmental knowledge graph, and mapping triple predicates into directed edges between nodes; the nodes in the cross-departmental knowledge graph include indicator nodes, process nodes, and decision nodes;

[0020] Extracting attribute information from the first data set, adding attributes to the nodes according to the attribute information, and generating node attributes; the node attributes include indicator node attribute information, process node attribute information, and decision node attribute information; the indicator node attribute information includes indicator name, indicator definition, and responsible department;

[0021] Label the nodes in the cross-departmental knowledge graph with semantic types and add initial node weights to the nodes;

[0022] Label the semantic types of directed edges in the cross-departmental knowledge graph and add initial edge weights to the directed edges; optimize the cross-departmental knowledge graph.

[0023] Furthermore, the optimization of the cross-departmental knowledge graph includes:

[0024] Calculate the semantic similarity SI between nodes in the cross-departmental knowledge graph, and set the semantic similarity SI greater than the preset first similarity threshold θ 1 The nodes of are defined as synonymous concept nodes;

[0025] Add synonymous relationship edges between synonymous concept nodes;

[0026] Identify the upper and lower concept nodes in the cross-departmental knowledge graph, and add belonging relationship edges between the upper and lower concept nodes.

[0027] Furthermore, generating the indicator semantic conflict list includes:

[0028] Retrieve all indicator nodes in the cross-departmental knowledge graph and extract indicator node attribute information;

[0029] Vectorize the indicator name and indicator definition in the indicator node attribute information to form an indicator vector of the indicator corresponding to the indicator node;

[0030] Calculate the similarity S between the indicator vectors of each indicator 1 , set the second similarity threshold θ 2 , for the similarity S 1 Higher than θ2 However, the indicator pairs with different responsible departments are marked as semantically conflicting indicators and added to the indicator semantic conflict list.

[0031] Furthermore, the construction of the cross-departmental knowledge graph after semantic mapping includes:

[0032] For each indicator pair in the indicator semantic conflict list, query the preset semantic mapping rule library to match the mapping rule; the semantic mapping rule is expressed in the form of IF-THEN;

[0033] According to the matched mapping rules, the mapping relationship edge and mapping attributes between the two semantic conflicting indicators in the indicator pair are added to form a cross-departmental semantic mapping layer;

[0034] Integrate the cross-departmental semantic mapping layer into the cross-departmental knowledge graph to form a cross-departmental knowledge graph after semantic mapping.

[0035] Further, the second data set includes real-time business parameters and real-time environmental data across functional departments;

[0036] The fused cross-departmental knowledge graph and semantic relevance matrix include:

[0037] Based on the real-time business parameters across functional departments, a real-time business feature vector is formed;

[0038] Based on the real-time environmental data across functional departments, a real-time environmental feature vector is formed;

[0039] In the cross-departmental knowledge graph after semantic mapping, the real-time business feature vector and the real-time environment feature vector are integrated to form a fused cross-departmental knowledge graph;

[0040] Based on the integrated cross-departmental knowledge graph, the semantic relevance between departments is calculated to generate an N×N semantic relevance matrix, where N is the number of departments.

[0041] Furthermore, generating a cross-departmental collaborative knowledge graph includes:

[0042] Extract the semantic relevance between departments from the semantic relevance matrix, and for those departments whose semantic relevance is higher than the preset relevance threshold θ 3 The first weight adjustment coefficient k 1 To improve the weights of its related nodes and edges, the first weight adjustment coefficient k 1 It is directly proportional to the semantic relevance;

[0043] For semantic relevance less than or equal to the preset relevance threshold θ 3 The department pairs are adjusted by the second weight coefficient k 2 To reduce the weights of its related nodes and edges, the second weight adjustment coefficient k2 It is inversely proportional to the semantic relevance.

[0044] Furthermore, the generation of cross-departmental collaborative decision-making suggestions based on the cross-departmental collaborative knowledge graph includes:

[0045] Collect the operation time series data of nodes in the cross-departmental collaborative knowledge graph, and obtain the time series dependency features between nodes based on the operation time series data and the pre-built time series dependency feature model;

[0046] According to the nodes and edges of the cross-departmental collaborative knowledge graph and the pre-built semantic interaction feature model, the semantic interaction features between nodes are obtained;

[0047] Fuse the temporal dependency features and the semantic interaction features to form a fused feature vector;

[0048] Build a logic rule library and establish a logic rule reasoning engine;

[0049] The fused feature vector is input into the logic rule reasoning engine for reasoning to generate cross-departmental collaborative decision-making recommendations.

[0050] A reasoning system based on semantic association and logical rules of knowledge graph, which is used to implement the above-mentioned reasoning method based on semantic association and logical rules of knowledge graph, and the system includes:

[0051] Graph construction module: used to obtain a first data set across functional departments, construct a structured data table based on the first data set, and construct a cross-departmental knowledge graph based on the structured data table;

[0052] Graph optimization module: extract indicator nodes from the cross-departmental knowledge graph, identify semantic conflict indicators, and generate an indicator semantic conflict list; based on the indicator semantic conflict list, construct a semantically mapped cross-departmental knowledge graph; obtain a second data set across functional departments, and based on the second data set and the semantically mapped cross-departmental knowledge graph, obtain a fused cross-departmental knowledge graph and a semantic relevance matrix; extract the weights of each node and edge of the fused cross-departmental knowledge graph, and dynamically adjust the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix to generate a cross-departmental collaborative knowledge graph;

[0053] Decision reasoning module: Generate cross-departmental collaborative decision-making recommendations based on the cross-departmental collaborative knowledge graph.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention integrates the data of key business departments such as production, quality, and supply chain by constructing a cross-departmental knowledge graph, realizes comprehensive integration and correlation analysis of data, breaks down the data barriers between departments, and provides a unified semantic basis for cross-departmental collaboration. Identifying indicator semantic conflicts and constructing semantic mappings solves the problem of inconsistent definitions and understandings of the same business indicator by different departments, eliminates semantic ambiguity, ensures the consistency and comparability of cross-departmental data, and provides a reliable basis for collaborative decision-making. Introducing real-time business and environmental data, dynamically optimizing the knowledge graph, so that the knowledge graph can reflect the current status of business operations in real time, capture business changes, and quantify the correlation between businesses in different departments, and timely reveal collaborative risks and optimization opportunities. Integrating temporal dependency features and semantic interaction features, using logical reasoning to form cross-departmental collaborative decision-making recommendations. Fully tapping into domain knowledge and data value makes decision recommendations more comprehensive and reasonable, while giving knowledge graphs learning and reasoning capabilities to achieve a closed loop from data to decision-making. Compared with traditional cross-departmental collaboration methods, the method provided by the present invention can significantly improve collaboration efficiency, timely discover and resolve conflicts between departments, optimize resource allocation, and shorten decision-making cycles. The quality of collaboration is also guaranteed, and decision-making is more scientific, transparent, and explainable, avoiding subjective biases and blind spots. The innovative application of knowledge graphs and logical reasoning in this invention provides new ideas and methods for efficient collaboration in complex organizational environments, and has broad application prospects. It is not only applicable within the enterprise, but can also be extended to collaboration between enterprises and even the industrial chain, helping to improve organizational resilience and competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 It is a principle flow chart of a reasoning method based on semantic association and logical rules of knowledge graph in the present invention;

[0058] Figure 2 A flowchart of a method for constructing a structured data table in a reasoning method based on semantic association and logical rules of a knowledge graph of the present invention;

[0059] Figure 3 A flowchart of a method for constructing a cross-departmental knowledge graph in a reasoning method based on semantic association and logical rules of a knowledge graph according to the present invention;

[0060] Figure 4A flowchart of a method for optimizing a cross-departmental knowledge graph in a reasoning method based on semantic associations and logical rules of a knowledge graph according to the present invention;

[0061] Figure 5 A flowchart of a method for generating an indicator semantic conflict list in a reasoning method based on semantic association and logical rules of a knowledge graph of the present invention;

[0062] Figure 6 A flowchart of a method for constructing a cross-departmental knowledge graph after semantic mapping in a reasoning method based on semantic association and logical rules of a knowledge graph of the present invention;

[0063] Figure 7 A flowchart of a method for obtaining a fused cross-departmental knowledge graph and a semantic relevance matrix in a reasoning method based on semantic relevance and logical rules of a knowledge graph according to the present invention;

[0064] Figure 8 This is a functional module diagram of a reasoning system based on semantic associations and logical rules of a knowledge graph in the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] Example 1

[0067] See also Figure 1 As shown, this embodiment provides a reasoning method based on semantic association and logical rules of knowledge graph, including:

[0068] Step S1000, obtaining a first data set across functional departments, constructing a structured data table according to the first data set, and constructing a cross-departmental knowledge graph based on the structured data table;

[0069] Further, step S1000 includes:

[0070] Step S1100, obtaining a first data set of cross-functional departments, wherein the first data set includes indicator definition documents, business process description documents, and historical decision records of cross-functional departments; the cross-functional departments include a production department, a quality management department, and a supply chain department;

[0071] Specifically, cross-functional departments refer to departments that have different functions within an organization but are interrelated and have impact on each other in terms of business. For example, the production department is mainly responsible for product manufacturing, the quality management department is responsible for quality control, and the supply chain department is responsible for raw material procurement and finished product shipment. They work closely together in production operations. To obtain data from these departments, you need to submit a data application to the relevant departments.

[0072] Initiate a data application to the production department to obtain definition documents of production-related indicators (such as output, qualified rate, etc.), business process description documents (such as production planning process, process flow, etc.), and historical production decision-making meeting minutes. Among them, output refers to the number of qualified products produced in a certain period, which is an important indicator to measure production capacity; qualified rate refers to the percentage of qualified products produced in a certain period to the total output, reflecting the product quality level. The production planning process stipulates the steps and requirements for formulating production plans, and the process flow describes in detail the operating specifications of each production process. Historical production decision meetings record major decisions made in production management, such as capacity adjustment, process improvement, etc. By obtaining these data, you can fully understand the production operation status and management measures. This helps to identify risks and bottlenecks in production and optimize resource allocation.

[0073] Initiate a data application to the quality management department to obtain definition documents of quality-related indicators (such as rework rate, customer complaint rate, etc.), quality management process documents (such as quality inspection process, quality audit process, etc.), and quality accident handling decision records. Among them, the rework rate refers to the percentage of the number of products that are reworked and repaired in the total output, reflecting quality problems in the production process; the customer complaint rate refers to the percentage of customer complaints in the sales volume, reflecting the quality of product use. The quality inspection process stipulates the quality inspection methods and standards for raw materials, semi-finished products, and finished products, and the quality audit process is used to identify non-conformities in the quality management system. Quality accident handling records the response measures to major quality problems, such as product recalls and compensation. Obtaining data from the quality management department can identify key factors that affect product quality and improve the quality control mechanism.

[0074] Initiate a data application to the supply chain department to obtain definition documents of delivery-related indicators (such as delivery timeliness rate, inventory turnover rate, etc.), supply chain management process documents (such as procurement process, logistics management process, etc.), and supplier selection decision records. The delivery timeliness rate refers to the percentage of orders delivered on time to the total number of orders. The inventory turnover rate refers to the ratio of product sales cost to the average inventory balance in a certain period of time, reflecting the efficiency of inventory management. The procurement process includes management regulations for supplier selection, procurement planning, order execution and other links. The logistics management process involves operational instructions for transportation, warehousing, distribution and other aspects. The supplier selection record reflects the factors considered in supplier management decisions, such as quality, cost, delivery time, etc. Supply chain data helps to optimize procurement strategies and inventory control and improve the overall performance of the supply chain.

[0075] Conduct integrity checks on the data sets obtained from various departments, identify missing or inconsistent data items, form data quality reports, and send them to relevant departments for supplementation and correction. Missing data may be caused by non-registration, omissions, etc., and inconsistent data may be caused by misunderstanding of the caliber or changes in statistical caliber. Timely identification and correction of data quality issues can provide a reliable foundation for subsequent data applications.

[0076] Step S1200, constructing a structured data table according to the first data set;

[0077] Furthermore, if Figure 2 As shown, step S1200 includes:

[0078] Step S1210, extracting the indicator name A, the first relationship type and the indicator name B from the indicator definition document to form an indicator information triple; wherein the indicator name A is the subject of the indicator information triple, the first relationship type is the predicate of the indicator information triple, and the indicator name B is the object of the indicator information triple; the first relationship type includes a calculation relationship, an influence relationship and a composition relationship;

[0079] Step S1220, extracting the process node A', the second relationship type and the process node B' from the business process description document to form a process triple; wherein the process node A' is the subject of the process triple, the second relationship type is the predicate of the process triple, and the process node B' is the object of the process triple; the second relationship type includes a temporal relationship and a logical relationship;

[0080] Step S1230, extracting decision events, third relationship types, decision problems or solutions from historical decision records to form decision information triples; wherein the decision event is the subject of the decision information triple, the third relationship type is the predicate of the decision information triple, and the decision problem or solution is the object of the decision information triple; the third relationship type includes the relationship between the decision event and the decision problem, and the relationship between the decision event and the decision solution;

[0081] Step S1240, constructing a structured data table based on the indicator information triples, the process triples and the decision information triples, the structured data table including triple subjects, triple predicates and triple objects.

[0082] Specifically, the indicator information triple indicates that there is a certain semantic association between two indicators, such as one indicator can be calculated from another indicator, or one indicator affects the value of another indicator, or one indicator is composed of multiple subordinate indicators. Taking "<output, calculated from, number of qualified products>" as an example, the output indicator can be calculated from the number of qualified products, and there is a calculation relationship between the two. Constructing the indicator information triple is conducive to describing the semantic relationship between indicators and laying the foundation for constructing the indicator system. The indicator system is built on the indicator information triple, which quantitatively describes the performance of the enterprise in production, quality, delivery, etc., and supports performance appraisal and improvement.

[0083] Business process description documents usually describe the execution steps and sequence of business activities in the form of natural language text. Using natural language processing technology, through syntactic dependency analysis, process node A', second relationship type, and process node B' can be extracted to form a process triple. The process node here refers to the activities in the business process, and the second relationship type describes the temporal dependency or logical relationship between activities. Taking "<workpiece cleaning, follow-up, workpiece assembly>" as an example, the workpiece assembly is arranged after the workpiece cleaning, and there is a temporal sequence between the two. Another example is "<parts processing, parallel, part inspection>" describes that two activities can be carried out at the same time without dependency constraints. The process triple reflects the temporal logic of business activities and is an important basis for process optimization and monitoring. The process path network is constructed through the process triple, and the bottleneck process is analyzed and resource allocation is optimized in combination with time parameters. When the actual process execution deviates from the temporal logic set by the process triple, timely warning helps process control.

[0084] Historical decision records carry the experience of enterprises in dealing with challenges in production, quality, supply, etc., and contain valuable management wisdom. Decision events, decision problems, decision solutions, and the relationship between them are extracted from decision records to form decision information triples. Taking "<Decision event 1, targeting, high repair rate>" as an example, it is revealed that in "Decision event 1", the problem of "high repair rate" needs to be solved. Furthermore, "<Decision event 1, adopt, increase the frequency of quality inspection>" reveals that in response to the problem of "high repair rate", the measure taken in "Decision event 1" is "increase the frequency of quality inspection". Decision information triples connect decision scenarios, decision problems, and solutions, presenting a paradigm of empirical knowledge. Decision information triples can be used for retrieval, recommendation, and reuse of decision knowledge, and can also be used as the basis for case reasoning. When faced with similar decision scenarios, reference solutions can be quickly matched based on decision information triples to assist decision makers in making judgments.

[0085] After extracting the indicators, processes, and decision-making data scattered in different departments into a unified triple form, a structured data table is constructed with subject, predicate, and object as fields. Each triple forms a row of records in the data table. Taking the indicator information triple as an example, a triple "<output, calculated from, number of qualified products>" can form a row of records in the data table, where "output" is the subject, "calculated from" is the predicate, and "number of qualified products" is the object. Constructing a structured data table based on triples eliminates the heterogeneity of business data between departments in terms of data organization, and has a unified structure of triples of subject, predicate, and object. This unity is conducive to the subsequent indexing, querying, and integrated analysis of relationships of data table fields. At the same time, the triple structure has clear semantics and is easy to interpret manually. Relying on the structured data table, the influence transmission between key indicators can be discovered, the optimization space in the process path can be explored, and the formation mechanism of the decision plan can be traced. The triple structure enables cross-departmental business data to be uniformly included in the analysis field and gain insight into business collaboration.

[0086] Step S1300, constructing a cross-departmental knowledge graph based on the structured data table.

[0087] Furthermore, if Figure 3 As shown, step S1300 includes:

[0088] Step S1310, mapping triple subjects and triple objects into nodes in a cross-departmental knowledge graph, and mapping triple predicates into directed edges between nodes; the nodes in the cross-departmental knowledge graph include indicator nodes, process nodes, and decision nodes;

[0089] Specifically, building a cross-departmental knowledge graph is to convert the triple information in the structured data table into a graph data structure. The subject and object of the triple are mapped to the nodes of the graph, and the predicate is mapped to the directed edges between the nodes. For example:

[0090] <Production, calculation formula, number of qualified products>

[0091] The subject "output" and the object "number of qualified products" are both indicator nodes, and the predicate "calculation formula" is the relationship type between them. This means that in the knowledge graph, there is a semantic association of "calculation formula" between the "output" node and the "number of qualified products" node. Indicator nodes represent the performance indicators that companies are concerned about. The semantic network between indicators is established through directed edges, which is conducive to insight into the indicator system.

[0092] <Workpiece cleaning, follow-up, workpiece assembly>

[0093] <Workpiece assembly, follow-up, quality inspection>

[0094] Here, the subject and object are both process nodes, and the predicate is "successor", which indicates the order of the process. In the knowledge graph, a path is formed from the "workpiece cleaning" node to the "workpiece assembly" node, and the "workpiece assembly" node points to the "quality inspection" node. The process nodes and their temporal relationships reflect the execution logic of business activities and are an important basis for process optimization and monitoring.

[0095] <Decision event 1, targeting, high repair rate>

[0096] <Decision event 1, adopt, increase the frequency of quality inspection>

[0097] These two triples share the subject "decision event 1" and correspond to the same decision node in the knowledge graph. There is a "targeting" relationship between the "decision event 1" node and the "high repair rate" node, and an "adopting" relationship between the "decision event 1" node and the "increase quality inspection frequency" node. Decision nodes record decision scenarios and response measures, which facilitates the reuse of decision knowledge.

[0098] Map structured data into a knowledge graph so that indicators, processes, and decision-making information scattered across different business links can be associated in the form of graph data. Different business elements are identified by the node types of the graph, and the direction of the edge reflects the business semantics, providing a unified knowledge base for cross-departmental collaborative optimization. When constructing a knowledge graph, nodes with the same name need to be merged. If different triple subjects or objects refer to the same business entity, they should be merged into the same node in the graph instead of creating multiple nodes repeatedly. The merging of nodes with the same name can improve the simplicity and connectivity of knowledge representation and reduce redundancy and ambiguity.

[0099] Step S1320, extracting attribute information from the first data set, adding attributes to the nodes according to the attribute information, and generating node attributes; the node attributes include indicator node attribute information, process node attribute information, and decision node attribute information; the indicator node attribute information includes indicator name, indicator definition, and responsible department;

[0100] Specifically, when constructing knowledge graph nodes, it is necessary to enrich the attribute information of the nodes so that they carry more semantic connotations. The attributes of different types of nodes have different focuses:

[0101] The attribute information of the indicator node includes: indicator name, indicator definition, responsible department, etc. The indicator name is the identifier of the indicator, the indicator definition explains the meaning of the indicator, the calculation formula, etc., and the responsible department clarifies the assessment of the indicator. For example, the attributes of the node "output" may be <name, output>, <definition, the number of qualified products produced in a certain period>, <responsible department, production department>. The addition of indicator attributes makes the node clear and can answer questions such as "what is this indicator, how to calculate it, and who is responsible".

[0102] The attribute information of a process node includes: process name, process description, department, etc. The process name identifies a business activity, the process description explains the input, conversion, output, etc. of the activity, and the department defines the division of responsibilities for process management. For example, the attributes of the node "workpiece assembly" may be <name, workpiece assembly>, <description, assemble the processed workpiece parts into a whole according to the drawing requirements>, <department, production workshop>. The process attributes enrich the connotation of the process node and enhance the information basis for process analysis and improvement.

[0103] The attribute information of the decision node includes: decision event name, decision problem, decision plan, responsible department, etc. The decision event name identifies a decision behavior, the decision problem explains the challenges faced, the decision plan summarizes the response measures, and the responsible department defines the decision participants. For example, the attributes of the "Decision Event 1" node may be <Event Name, 2022 First Quarter Production Meeting>, <Decision Problem, Rework Rate Exceeds Standard>, <Decision Plan, Strengthen First Inspection, Adjust Process Parameters>, <Responsible Department, Quality Department, Production Department>. Decision attributes provide more comprehensive contextual information and enhance the pertinence of decision traceability and review.

[0104] Node attributes are an important component of the knowledge graph, and are a semantic description of the node's connotation. The richer the attributes, the richer the knowledge contained in the node. The attribute information contained in the business data set is extracted into node attributes and organized uniformly in the knowledge graph, so that the scattered and unstructured business knowledge can be integrated and structured, thus serving the unified semantic association analysis and comprehensive application.

[0105] Step S1330, annotating the nodes in the cross-departmental knowledge graph with semantic types and adding initial node weights to the nodes;

[0106] Specifically, the nodes in the knowledge graph may belong to different semantic types. By labeling the types of nodes, their roles in the business semantic network can be clarified. Common semantic types include indicators, processes, decisions, etc., and types can also be customized according to business needs. During the graph construction process, the semantic type of each node needs to be judged and labeled. For example, the node "output" is labeled as <node type, indicator>, the node "artifact assembly" is labeled as <node type, process step>, and the node "decision event 1" is labeled as <node type, decision>. Through semantic type labeling, the knowledge graph presents a global classification view, making knowledge organization more organized. Type-based graph indexing and query will also be more efficient.

[0107] In addition to type annotation, it is also necessary to add initial weights to the nodes. The node weight reflects the importance of the node and can be assigned based on the node's attributes. For example, the weight of the indicator node can be derived from the score weight of the indicator in the performance appraisal, the weight of the process node can be derived from the proportion of working hours in the process steps, and the weight of the decision node can be derived from the impact assessment of the decision. The initial weight assignment can be based on objective data or comprehensive expert experience. By weighting the nodes, the knowledge graph forms a distribution map of key knowledge and guides the focus of the analysis. Subsequently, the connection relationship between the nodes can be combined, and algorithms such as PageRank can be used to iteratively optimize the node weight values ​​to mine the key nodes in the graph.

[0108] Step S1340, annotating the semantic types of the directed edges in the cross-departmental knowledge graph and adding initial edge weights to the directed edges;

[0109] Specifically, the directed edges between nodes in the knowledge graph represent different types of relationships. Semantic annotation of relationship types helps to clarify the semantic connections between business elements. For example, there may be mathematical operation relationships such as <addition> and <division> between indicator nodes, there may be temporal dependency relationships such as <precedes> and <successors> between process nodes, and there may be influence relationships such as <promotion> and <constraint> between indicators and processes. By predefining these semantic relationship types and annotating the relationship type for each edge in the graph, business knowledge can be organized in a more flexible and expressive way. This provides a basis for complex semantic association analysis, such as multi-hop relationship queries (such as "a <influences> b <influences> c", etc.) and pattern matching (such as "process 1 <precedes> process 2 <precedes> process 3", etc.).

[0110] While performing semantic type annotation, it is also necessary to add weight attributes to directed edges to indicate the strength of the relationship. The initial edge weight can be assigned based on the confidence, frequency, expert rating, etc. of the relationship, such as the number of data dependencies between two indicators, the business relevance of two process steps, etc. Weighted edges enable the knowledge graph to represent the strength of business associations, enhancing its application flexibility. For example, in a risk analysis scenario, node associations with high edge weights represent possible transmission paths for risk impacts; in a knowledge recommendation scenario, edges with high weights correspond to more relevant node combinations. Subsequently, multiple connection paths between nodes can be used to evaluate the structural weights and semantic weights of edges, and dynamically optimize edge weight values.

[0111] Step S1350, optimizing the cross-departmental knowledge graph.

[0112] Furthermore, if Figure 4 As shown, step S1350 includes:

[0113] Step S1351, calculate the semantic similarity SI between nodes in the cross-department knowledge graph, and set the semantic similarity SI greater than the preset first similarity threshold θ 1 The nodes of are defined as synonymous concept nodes;

[0114] The calculation of the semantic similarity SI between nodes in the cross-departmental knowledge graph includes:

[0115]

[0116] in:

[0117] Representation Node and nodes The semantic similarity between The larger the value, the more similar the semantics between nodes.

[0118] and Respectively represent nodes and nodes The embedded vector representation of can be learned through a knowledge graph embedding model (such as TransE).

[0119] represents the cosine similarity of two embedding vectors and is defined as:

[0120]

[0121] and Respectively represent nodes and nodes A collection of attributes.

[0122] The Jaccard similarity coefficient between two attribute sets is defined as:

[0123]

[0124] Representation Node and The first Relationship characteristic function , used to capture the structural information of nodes in the graph, is the number of relationship features between nodes.

[0125] In a cross-departmental knowledge graph, there may be multiple structural relationships between different nodes, such as:

[0126] Shortest path length: The shortest path between nodes reflects their distance in the graph;

[0127] Number of common neighbors: The number of direct neighbors that nodes share can indicate their similarity;

[0128] Path diversity: the number of paths between nodes connected by different types of edges;

[0129] Information flow weight: The weight or influence of information transmitted between nodes through a specific path.

[0130] Each relationship can be defined as a characteristic function ,pass To identify the number of these features.

[0131] effect:

[0132] Capture the structure information of the node: The size of determines the type and granularity of the relationship features, and can characterize the connection relationship between nodes from multiple angles.

[0133] Weighing complexity and precision: increasing This can improve the accuracy of similarity calculation, but will increase the computational complexity. Therefore, in practical applications, it is necessary to select an appropriate number of features based on the scenario.

[0134] For the The weight coefficient of the relationship feature.

[0135] are the weight factors of embedding similarity, attribute similarity and relationship features, satisfying Their values ​​can be set empirically or optimized using parameter search methods such as grid search.

[0136] This formula comprehensively considers the semantic representation of nodes, attribute similarity, and structural relationships in the knowledge graph, and more comprehensively describes the semantic similarity between nodes. The setting of parameters requires a balance between computational efficiency and effect. With the improvement of the similarity of node embedding vectors, the increase of attribute overlap, and the enhancement of relationship characteristics, the semantic similarity SI between nodes will also increase accordingly. This similarity formula can help identify semantically equivalent or highly related nodes in cross-departmental knowledge graphs, provide a basis for the subsequent construction of mapping relationships, elimination of redundancy, and fusion of knowledge, and improve the semantic consistency and compactness of knowledge graphs.

[0137] Step S1352, adding synonymous relationship edges between synonymous concept nodes;

[0138] Step S1353, identify the upper and lower concept nodes in the cross-departmental knowledge graph, and add belonging relationship edges between the upper and lower concept nodes.

[0139] Specifically, synonymous concept nodes in the knowledge graph are identified by calculating semantic similarity. Semantic similarity comprehensively considers multiple features of the node, including:

[0140] (1) Embedded vector representation of nodes. Using knowledge graph embedding models such as TransE and TransR, we can learn a low-dimensional dense vector representation of each node. The similarity of vectors (such as cosine similarity) can characterize the semantic similarity of nodes in the embedding space. For example, the two nodes "product" and "commodity" have similar neighbor nodes and connection relationships in the graph, so the learned embedding vectors are also relatively close.

[0141] (2) Node attribute characteristics. Directly compare the attribute sets of two nodes. If the attribute overlap is high (such as the Jaccard similarity coefficient), it indicates that the two nodes have strong similarity in attribute semantics. For example, the two indicators "shipment volume" and "sales volume" may have highly consistent attributes such as definition description and calculation formula.

[0142] (3) Structural features of nodes in the graph. The structural relationship of nodes in the knowledge graph can also represent its semantics. For example, two nodes with multiple common neighbor nodes have a higher semantic association. For example, if there are many connection paths between two nodes (such as "A'' produces B'', B'' transports C'', A'' sells C''"), their relevance can also be inferred. "Web pages" and "articles" may play similar structural roles in the knowledge graph.

[0143] When calculating semantic similarity, it is necessary to comprehensively weigh the above factors, and the final similarity score of the node pair can be obtained by weighted average. In order to automatically identify synonymous concepts, a threshold needs to be set, and node pairs above this threshold can be defined as synonymous concepts. The threshold setting needs to balance the accuracy and recall rate, which can be calibrated through experimental statistics and expert evaluation.

[0144] Identifying synonymous concepts in the knowledge graph can consolidate differences in the expression of business concepts and improve the standardization of knowledge organization. For example, "user" and "customer" may have different expressions in different business departments, but they refer to the same concept semantically and need to be unified. When building a cross-departmental knowledge graph, it is especially necessary to deal with the problem of name ambiguity when cross-referencing concepts between departments. Synonymous concept alignment based on semantic similarity is an effective method.

[0145] After the synonymous concept nodes are identified in step S1351, it is necessary to explicitly add "synonymous relationship" edges between these nodes to form a special edge type. This explicit modeling facilitates the association analysis and query of synonymous concepts, so that the knowledge graph can easily answer questions such as "which concepts are synonymous". Integrate the synonymous relationship into the structure of the knowledge graph in the form of an edge, making it a knowledge element that can be reasoned about. For example, through the connection of synonymous relationship edges, nodes such as "article-synonym->webpage", "webpage-synonym->document", and "document-synonym->material" form a "concept family" to represent their semantic relevance.

[0146] Unifying the representation of synonymous concepts helps simplify the knowledge graph and reduce potential redundancy and inconsistency. By adding synonymous relationship edges, the "equivalence class" of concepts can be flexibly established without destroying the original graph structure, which not only retains the diverse expressions of concepts but also achieves the logical unification of concepts. Synonymous relationship edges can assist knowledge reasoning, such as extended retrieval based on synonymous concepts (such as when searching for "articles", synonymous "web pages" and "documents" are also returned), so that the knowledge graph can better fit the diverse language usage habits of users.

[0147] In addition to adding synonymous edges, redundant nodes also need to be merged when optimizing the knowledge graph. For nodes with extremely high semantic similarity (such as similarity equal to 1), from the perspective of simplifying the graph, only one of the nodes can be retained and its neighboring edges can be merged to simplify the knowledge representation. At the same time, the "synonymous concept" attribute should be added to the merged node, and all synonymous concept names should be listed to ensure that the diversity information of the concept representation is not lost. For example, if "product" and "commodity" are identified as completely synonymous, only the "product" node can be retained, and all edges pointing to "commodity" can be connected to "product", and the <synonymous concept, "commodity"> attribute should be added to the "product" node.

[0148] When merging synonymous concept nodes, it is necessary to consider updating the weights. The weights of the synonymous concept nodes can be accumulated to form a new weight for the merged node. The merging of weights allows the importance of synonymous concept nodes to be preserved in the simplified graph. For example, after the "user" and "customer" nodes are merged, the weight of the new node is the sum of the weights of the two, reflecting the overall weight of the "user / customer" concept in the entire graph.

[0149] The hierarchical concepts reflect the hierarchical semantic relationship of concepts in the knowledge graph. By identifying the hierarchical associations of nodes and building a conceptual hierarchy, the systematic nature of knowledge organization can be improved and the ability of knowledge reasoning can be enhanced. For example, the two nodes "output" and "production index" are conceptually reflected in the hierarchical relationship of "output is a production index".

[0150] The identification of hyponyms and hyponyms can be achieved by combining the following strategies:

[0151] (1) Using the ontology knowledge base. Map the hierarchical relationships solidified in the domain ontology knowledge base to the corresponding nodes of the knowledge graph, and obtain a batch of hierarchical relationship seeds in advance.

[0152] (2) Graph mining. Analyze the structural pattern of the knowledge graph. If there are many ISA (“is a kind of”) type edges between two nodes, they tend to be in a hierarchical relationship, such as “output-ISA->production index” and “qualified rate-ISA->production index”. The symbol “->” indicates a “relationship” or “direction”.

[0153] (3) Distributed semantics. Compare the word embedding vectors of the words corresponding to the two nodes. If there is a certain vector difference between the hypernym embedding and the hyponym embedding, such as hypernym embedding - hyponym embedding ≈ constant vector, then the hypernym relationship can be inferred, such as vec("production index") - vec("output") ≈ vec("quality index") - vec("repair rate").

[0154] (4) Pattern matching. Match the expression templates of the hypernymy relationship such as "E is a kind of F" and "F includes E" in a large-scale text corpus, count the co-occurrence frequency, and discover the hypernymy relationship of the concepts based on this.

[0155] By using a variety of strategies in combination, we can more comprehensively mine the semantic relationships between the upper and lower levels contained in the knowledge graph. In order to reflect the hierarchical organization of concepts, it is necessary to add "belongs to" relationship edges to the identified hierarchical and lower level concept nodes to form a directed semantic hierarchy of concepts. The "belongs to" edge points from the lower level concept to the higher level concept, expressing the semantics of "the lower level belongs to the higher level" and "the higher level contains the lower level". For example, "output-belongs to->production indicators" and "production indicators-belongs to->performance indicators" form a hierarchical and lower level semantic chain of "output <production indicators <performance indicators".

[0156] The hierarchical hierarchy of upper and lower concepts formed based on the "belongs to" relationship edge enables the knowledge graph to have the ability of conceptual abstraction and concept refinement. Abstract generalization can be achieved through the upper concept, such as the abstraction of "output" into "production indicators"; concretization can be achieved through the lower concept, such as the refinement of "production indicators" into "output". During query and reasoning, the semantic extension and connotation capabilities of this concept make the knowledge graph more intelligent and closer to the way people understand. For example, when asking "How is the completion of production indicators in the first quarter of 2022?", the knowledge graph can use the "belongs to" edge to summarize the statistical data of lower concepts such as "output" and "qualified rate" to form an overall performance of "production indicators".

[0157] The weights of superordinate and subordinate concepts need to be reconsidered based on the edges. From a semantic perspective, the "belongs to" edge expresses a "belongs to" relationship, and the weight of the superordinate concept should be higher than the sum of the weights of the direct subordinate concepts. Therefore, it is possible to consider using an additive model to update the weights of the superordinate and subordinate nodes. For example, if node E "belongs to" node F, the weight of F is equal to the sum of the original weight of F and the weight of E. This weight propagation based on the "belongs to" edge can objectively reflect the cumulative importance of concepts in the semantic network. For example, the sum of the weights of "output" and "qualified rate" is transferred to the superordinate concept "production indicators", making the overall weight of "production indicators" exceed any of its subordinate concepts. The weight allocation in the concept hierarchy helps to prioritize knowledge reasoning, such as analyzing problems from superordinate concepts with high weights.

[0158] In general, the three sub-steps of knowledge graph optimization in step S1350 enrich the semantic association and organization of the knowledge graph from different levels. Synonymous relationships reveal the equivalence of concepts and simplify the graph; hierarchical relationships reveal the hierarchy of concepts and systematic organization; the attributes and weights of nodes and edges further characterize the semantics of elements. The integration of the three aspects of optimization makes the cross-departmental knowledge graph a highly semantic business knowledge representation, laying a solid foundation for intelligent analysis and decision-making.

[0159] Step S2000, extracting indicator nodes from the cross-departmental knowledge graph, identifying semantic conflict indicators, and generating an indicator semantic conflict list; constructing a semantically mapped cross-departmental knowledge graph based on the indicator semantic conflict list; obtaining a second data set across functional departments, and obtaining a fused cross-departmental knowledge graph and a semantic relevance matrix based on the second data set and the semantically mapped cross-departmental knowledge graph; extracting the weights of each node and edge of the fused cross-departmental knowledge graph, and dynamically adjusting the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix to generate a cross-departmental collaborative knowledge graph;

[0160] Furthermore, step S2000 includes:

[0161] Step S2100, extracting indicator nodes from the cross-departmental knowledge graph, identifying semantic conflict indicators through semantic similarity calculation, and generating an indicator semantic conflict list;

[0162] Furthermore, if Figure 5 As shown, step S2100 includes:

[0163] Step S2110, retrieving all indicator nodes in the cross-departmental knowledge graph and extracting indicator node attribute information;

[0164] Step S2120, vectorizing the indicator name and indicator definition in the indicator node attribute information to form an indicator vector of the indicator corresponding to the indicator node;

[0165] Step S2130, calculate the similarity S between the indicator vectors of each indicator 1 , set the second similarity threshold θ 2 , for the similarity S 1 Higher than θ 2 However, the indicator pairs with different responsible departments are marked as semantically conflicting indicators and added to the indicator semantic conflict list.

[0166] Specifically, the indicator nodes in the cross-departmental knowledge graph carry the performance indicator information used by each department, including attributes such as indicator name, indicator definition, and responsible department. The indicator name reflects the popular expression of the indicator, the indicator definition explains the connotation of the indicator from a business perspective, and the responsible department clarifies the assessment object of the indicator. These attribute information are an important basis for conducting indicator semantic comparison and identifying semantic conflicts. Therefore, the first step is to traverse the graph, find all nodes of type "indicator", and extract the attribute information of the nodes. This process can be implemented with the help of the query language of the graph database. Taking the graph database Neo4j as an example, Cypher statements can be used for retrieval and extraction. Cypher is a declarative graph query language with a simple structure similar to SQL.

[0167] Through the query of the graph database, the indicator node information in the knowledge graph can be quickly retrieved and extracted, making full use of the advantages of the graph in the structured representation of business knowledge. The node-edge-attribute model of the graph structure intuitively organizes the indicator information scattered in different business fields. The Cypher query language expresses the retrieval logic in a graph manner, without the need to split complex connection operations, which greatly simplifies the process of extracting cross-departmental indicator information. At the same time, the graph database also supports rich indexes for nodes, edges, and attributes, thereby accelerating the query response of large-scale graphs.

[0168] Converting unstructured text data into numerical forms that can be understood and processed by computers is an important foundation for natural language processing. Word vector is a widely used text representation learning method that maps words to multidimensional real number space to form a word embedding. In the word vector space, semantically similar words are geometrically closer. Representing indicator information as word vectors can quantitatively characterize the semantic features of indicator names and definitions, and then compare indicator semantics through vector operations. Google's word vector model (Word2Vec) and Stanford's global vector model (GloVe) can be used to vectorize the indicator names and indicator definitions in the indicator information list. Vectorizing indicator information and embedding unstructured indicator text into a structured vector space is a key step in realizing indicator semantic association analysis. Word vectors convert complex language expressions into standardized mathematical forms, so that semantic similarity can be measured by vector distance, and semantic operations can be implemented with the help of vector algebra. At the same time, the pre-trained word vector model trained on massive text corpus well captures the semantic information with a wide range of vocabulary coverage, making the semantic representation ability of indicator vectors more comprehensive and accurate.

[0169] In step S2120, the name, definition and other attribute information of each indicator node has been vectorized to form a corresponding indicator vector. The indicator vector describes the semantic characteristics of the indicator with multi-dimensional real numbers. Intuitively speaking, in the vector space, the closer the geometric distance between two indicator vectors, the more similar the semantics of the indicators they represent. Therefore, by calculating the similarity between indicator vectors, the degree of semantic proximity of different indicators can be determined.

[0170] By building a cross-departmental knowledge graph and applying technical means such as word vectors and similarity thresholds on its basis, potential semantic conflict indicators within the enterprise are automatically discovered. These conflicts reflect the differences between different departments in data standards and are an important hidden danger that affects data quality and data value. Traditional manual combing is difficult to fully identify such problems. Knowledge graphs combined with semantic analysis technology provide an intelligent and efficient solution. For the identified semantic conflict indicators, we can summarize them to form a list of problems, and then carry out data governance. Through communication and coordination, we can unify the caliber, standardize the definition, eliminate differences, and finally form a unified indicator standard at the enterprise level. This is of great significance for improving the value of data applications and supporting scientific decision-making.

[0171] The similarity S between the index vectors of each index is calculated 1 include:

[0172]

[0173] in:

[0174] Indicators and indicators The similarity between the indicator vectors of The larger the value, the more similar the semantics of the indicators are.

[0175] and Respectively indicate indicators and indicators In the Here we assume that the indicator vector is represented by It consists of semantic dimensions.

[0176] Indicates The weight of a semantic dimension reflects the importance of the dimension to the similarity calculation. It can be learned through expert knowledge or data-driven methods.

[0177] Represents two vectors in The cosine similarity in dimensions is used to measure the consistency in the direction of the vector.

[0178] It is a regularization parameter that controls the effect of vector modulus difference on similarity. The larger it is, the greater the penalty for the difference in vector modulus length. It can be set empirically or optimized through parameter search.

[0179] Indicators and indicators In the The square of the Euclidean distance on the semantic dimension measures the numerical proximity of the vectors.

[0180] Indicators and indicators The attribute collection and The Jaccard similarity coefficient is calculated as follows:

[0181]

[0182] It measures the similarity between two indicators at the attribute level.

[0183] is the control parameter of the steepness of the Sigmoid function, is the control parameter of the center position of the Sigmoid function, and Together they determine the contribution of attribute similarity to the overall similarity. and It can be set based on business understanding or obtained through parameter learning.

[0184] This formula comprehensively considers the directional similarity, module length difference and attribute matching of indicator vectors, and fully characterizes the semantic correlation between indicators.

[0185] Introducing dimension weights , the influence of different semantic dimensions on similarity calculation can be flexibly adjusted.

[0186] Using the exponential function The vector modulus length differences are smoothed to reduce the impact of outliers.

[0187] Using Sigmoid function Normalize the attribute similarity and pass the parameter and Controls its contribution to the overall similarity.

[0188] When the vector directions of two indicators are more consistent, the difference in modulus length is smaller, and the attribute matching degree is higher, their similarity is On the contrary, the similarity will decrease.

[0189] By adjusting the parameters , the similarity calculation formula can be optimized to better adapt to specific fields and application scenarios. This helps to accurately identify semantic conflict indicators and generate a high-quality indicator semantic conflict list, providing a reliable basis for subsequent semantic mapping and knowledge fusion. At the same time, the calculation result of the similarity formula can also be used as the weight of the edge between indicator nodes to enrich the semantic representation ability of the knowledge graph. In short, this formula designs a robust and interpretable indicator similarity measurement method from multiple perspectives such as vector semantics, numerical features, and attribute matching.

[0190] Step S2200, traversing the indicator semantic conflict list, generating a unified semantic representation for each semantic conflict indicator, and constructing a cross-departmental knowledge graph after semantic mapping;

[0191] Furthermore, if Figure 6 As shown, step S2200 includes:

[0192] Step S2210, for each indicator pair in the indicator semantic conflict list, query a preset semantic mapping rule library to match the mapping rule; the semantic mapping rule is expressed in the form of IF-THEN;

[0193] Step S2220, according to the matched mapping rule, adding a mapping relationship edge and a mapping attribute between two semantically conflicting indicators in the indicator pair to form a cross-departmental semantic mapping layer;

[0194] Step S2230, integrating the cross-departmental semantic mapping layer into the cross-departmental knowledge graph to form a cross-departmental knowledge graph after semantic mapping.

[0195] Specifically, semantic conflict refers to different business departments using different terms to describe indicators of the same or similar concepts, or using the same terms but referring to different conceptual connotations. This phenomenon of semantic conflict is more common in cross-departmental collaboration and becomes a barrier to business interconnection. For example, the production department uses "output" to represent product output, while the finance department uses "output" to represent output value, and there is a difference in semantic reference between the two. For another example, although the terminology of the "qualified rate" of the production department and the "first inspection qualified rate" of the quality department are similar, the denominator of the former is the number of production, and the latter is the number of inspections, and the calculation caliber is inconsistent. Semantic conflicts make it difficult for cross-departmental information to be directly connected and circulated, affecting collaborative efficiency.

[0196] In order to eliminate the semantic gap, it is necessary to traverse each pair of semantic conflict indicators identified in the semantic conflict list and give them a unified and standardized semantic representation through semantic mapping. There are two ways to unify the semantic representation: one is to select a term from the conflict indicator pair as the standard expression, and the other is to create a new standard term. The choice depends on the semantic similarity of the conflict indicators. The result of semantic mapping is to replace the original multiple synonym indicators with a unified term and establish a mapping association between the indicators. This process outputs the knowledge graph after semantic mapping. In the graph, indicators of the same concept are integrated into nodes of the unified term, eliminating semantic redundancy; indicators of different concepts are distinguished by attribute definitions even if the terms are similar, eliminating semantic confusion; for indicators with similar semantics, semantic tracking is maintained through mapping relationship edges.

[0197] Through data analysis and expert experience summary, some common indicator semantic mapping rules can be summarized. These rules are included in the semantic mapping rule base to guide the resolution of semantic conflicts. The rule base organizes rules in the form of IF-THEN. The antecedent (IF) part determines the semantic relationship type of the indicator pair, and the consequent (THEN) part gives the corresponding mapping strategy. The semantic relationship type can be determined based on the dimensions of the concept's extension overlap, connotation similarity, etc., mainly including: equivalent relationship, implication relationship, intersection relationship, etc. For example:

[0198] IF the extension of indicator G completely overlaps with that of indicator H THEN select the indicator term with high word frequency as the standard expression;

[0199] This rule applies to situations where the terms have different expressions but refer to exactly the same objects. The disambiguation strategy is to choose a broader indicator term.

[0200] IF the extensions of indicator G and indicator H exist in an inclusive relationship THEN select the indicator term with a larger extension as the standard expression.

[0201] This rule applies to situations where one indicator concept covers another indicator concept, and the disambiguation strategy is to choose the term with a broader concept extension.

[0202] IF the semantic similarity between the connotation of indicator G and indicator H is higher than the threshold THEN a new standardized indicator term is constructed using the superordinate concept, and a "synonymous" relationship edge is established between indicator G / H and the newly constructed indicator.

[0203] This rule applies to situations where concepts have similar connotations but are difficult to express directly and uniformly using the original terms of G / H. The disambiguation strategy is to newly construct a superordinate concept indicator that covers the connotations of G / H.

[0204] The semantic mapping rule base is not static and can be dynamically updated according to business changes. During the rule matching process, new mapping rules are generated for semantic relationship types that cannot be covered and added to the rule base. The addition and improvement of rules make the processing of semantic mapping standardized and the degree of automation continuously improved.

[0205] Apply the semantic mapping rules to the indicator pairs, that is, establish mapping relationship edges with attributes between the conflicting indicators. The mapping relationship type can be "equivalence", "implication", "intersection", etc., which are consistent with the semantic relationship type matched in the rule base. The attributes of the mapping edge record the extraction rules, confidence and other meta-information of the mapping relationship. Aggregate the mapping relationship edges in each conflicting indicator pair to form a semantic mapping layer. The semantic mapping layer is a semantic association network interwoven between conflicting indicators, in which the mapping relationship clearly depicts the semantic similarities and differences of each indicator. Visualize the semantic mapping layer to intuitively show the semantic context of cross-departmental terminology.

[0206] For indicator pairs that cannot be matched to mapping rules, it means that the semantic relationship they contain exceeds the expression capacity of the existing rule base, and new rules need to be developed. Experts analyze the semantics of these indicator pairs, determine their mapping paths, and summarize and refine new mapping rules. The new rules are incorporated into the rule base and applied to subsequent mapping practices. The semantic mapping rule base is continuously enriched and improved during use to meet the semantic association needs of different field scenarios.

[0207] The mapping relationship edges and their attributes in the semantic mapping layer need to be integrated into the cross-departmental knowledge graph to replace the original semantic conflict nodes and relationships in the graph. The fusion process requires coordinating the semantic representation of the original knowledge graph and the semantic mapping layer:

[0208] Traverse the mapping relationship edges in the semantic mapping layer, and add corresponding mapping relationship edges to the cross-departmental knowledge graph for the nodes at both ends of the edge. At the same time, synchronize the attributes of the mapping edge to the knowledge graph. For example, there is an "equivalence" relationship edge in the mapping layer connecting "finished product rate" and "qualified rate", and the corresponding "equivalence" edge needs to be added to the graph.

[0209] Then, the original indicator nodes with the same name and meaning and the related edges in the knowledge graph are deleted, and the indicator nodes with unified semantic representation are retained. The edges connected with "finished product rate" and "qualified rate" in the graph are deleted, and only the "qualified rate" node is retained as the standard representation of the semantic concept.

[0210] For the case where there is a mapping relationship in the semantic mapping layer but no corresponding node in the knowledge graph, these nodes and edges are directly added to the knowledge graph. For example, in the mapping layer, "output" and "output value" form a "cross" relationship, but the "output value" node is missing in the graph, in this case, the "output value" node and the "cross" mapping edge are added.

[0211] Semantic mapping standardizes the indicator terms in the cross-departmental knowledge graph and connects related indicators through mapping relationship edges, eliminating redundancy and ambiguity while maintaining semantic richness and continuity. The mapped knowledge graph contains the association between unified semantics and original semantics, making knowledge organization more standardized and flexible. Semantic mapping alleviates the communication barriers across departments and lays the foundation for building a unified semantic system. On the premise of semantic consistency, the business knowledge of various departments can be integrated to achieve a global understanding of business activities and optimize decision-making. Taking "qualified rate" as an example, semantic mapping clarifies the difference between it and "first inspection qualified rate". The former reflects the quality of the final product, while the latter focuses on production process control, and the two form a progressive relationship. When the qualified rate of the final product decreases, the qualified rate of the first inspection in the upstream link can be traced back to find the source process that introduces defects. The knowledge association after semantic unification extends quality management from end detection to source control, realizing cross-departmental quality collaboration.

[0212] Step S2300, obtaining a second data set across functional departments, wherein the second data set includes real-time business parameters and real-time environment data across functional departments; obtaining a fused cross-departmental knowledge graph and a semantic relevance matrix based on the second data set and the cross-departmental knowledge graph after semantic mapping;

[0213] Furthermore, if Figure 7 As shown, step S2300 includes:

[0214] Step S2310, forming a real-time business feature vector according to the real-time business parameters across functional departments;

[0215] Step S2320, forming a real-time environment feature vector based on the real-time environment data across functional departments;

[0216] Step S2330, fusing the real-time business feature vector and the real-time environment feature vector in the semantically mapped cross-departmental knowledge graph to form a fused cross-departmental knowledge graph;

[0217] Step S2340, based on the fused cross-department knowledge graph, calculate the semantic relevance between departments and generate an N×N semantic relevance matrix, where N is the number of departments;

[0218] Step S2350, normalizing the semantic relevance matrix.

[0219] Specifically, the second data set refers to the data that reflects the latest business operation status of each department that is continuously collected after the knowledge graph is built. Unlike the historical data used in the construction phase, the second data set emphasizes real-time and aims to provide the latest business perspective for the knowledge graph. Real-time business parameters may include order volume, production plan execution progress, equipment operating parameters, quality inspection results, etc., which depict the immediate performance of business processes. Real-time environmental data may include upstream and downstream supply and demand dynamics, industry development trends, macroeconomic trends, etc., which reflect changes in the external ecology of the enterprise. To obtain the second data set, it is necessary to connect with the information system of the business department in real time and collect environmental data from external channels. During the data collection process, attention should be paid to the timeliness, completeness and accuracy of the data. By integrating the second data set into the cross-departmental knowledge graph after semantic mapping, the knowledge graph is closely integrated with the actual business, thus becoming an important basis for enterprises to implement business decisions.

[0220] The real-time business feature vector is a quantitative representation of the business operation status of each department. Each element in the vector corresponds to a business parameter, and the parameter value reflects the real-time performance of the business element. To construct a real-time business feature vector, we must first determine the dimension of the vector, that is, how many business parameters to include. Usually, the parameters to be included are selected based on the business importance and data availability of each parameter. Then, the real-time data interface of the business department is accessed to obtain the current value of the selected parameter. The real-time data interface here can be a manufacturing execution system (MES), a supply chain management system (SCM), a warehouse management system (WMS), etc. Interface access needs to handle issues such as identity authentication and authority control to ensure data security. For heterogeneous data returned by different interfaces, it is necessary to clean, convert, and uniformly represent them to form standardized parameter values. Finally, the real-time values ​​of each parameter are assembled into a vector in the agreed order, that is, a real-time business feature vector. When the business parameters change, the real-time business feature vector is also updated accordingly. Taking the production department as an example, the real-time business feature vector may contain elements such as "current capacity utilization", "qualified rate of the current shift", and "equipment start-up rate". Through real-time business feature vectors, scattered departmental business data can be aggregated into a unified feature representation, providing a basis for cross-departmental knowledge association.

[0221] The business operation status across functional departments is quantified into real-time business feature vectors, and a unified data representation oriented to semantic association analysis is established. Compared with the original business parameters, the real-time business feature vector has a fixed structure and consistent semantics, which reduces the difficulty of cross-departmental data association. Through real-time feature extraction, inter-departmental business data that was originally difficult to directly compare is mapped to the same vector space, which is computable at the semantic level. This feature representation simplifies the complexity of the original data and highlights the key semantic relationships between business elements. This lays a data foundation for the subsequent cross-departmental business comparison, impact analysis, and joint decision-making.

[0222] The real-time environmental feature vector is designed to describe the macro environment in which the enterprise is located. Environmental factors such as industry policies and market conditions often have a global impact on various business departments. By perceiving environmental changes, enterprises can respond in a timely manner and innovate to change. To construct a real-time environmental feature vector, it is necessary to first define the relevant environmental factors. This requires analysis from the perspectives of industry attributes, industrial chain position, and competitive situation. The selection of environmental factors should take into account both importance and availability, and factors with great impact on various departments and relatively easy data access should be preferred. Then, various external data sources are accessed to obtain the real-time values ​​of each environmental factor. External data sources can be industry associations, third-party research institutions, upstream and downstream enterprises in the industrial chain, etc. It is necessary to establish a long-term data sharing mechanism with the data source and clarify the scope and frequency of data use. At the same time, do a good job of data outlier processing, identify and correct unreasonable extreme values. Finally, the real-time environmental data is processed through feature engineering to extract and form a real-time environmental feature vector with fixed dimensions. Taking home appliance manufacturing companies as an example, real-time environmental features may include "real estate prosperity index", "consumer confidence index", "key raw material price index", etc. Real-time environmental feature vectors enable enterprise managers to examine internal management from an open perspective and proactively adapt to external changes.

[0223] By introducing real-time environmental feature vectors, the knowledge graph can gain the ability to perceive the macro environment, which is conducive to enterprises maintaining strategic focus under complex and changing external situations. Since environmental factors often have cross-impacts on various business departments, it is difficult for a single department to fully grasp them. The real-time environmental feature vector provides "nerve endings" for cross-departmental collaboration, enabling departments to form linkages based on consistent environmental cognition. When the real-time environmental feature vector indicates a major change in the external situation, the knowledge graph can quickly calculate the impact on each business element, thereby guiding relevant departments to adjust their decisions in a timely manner. The shift from passive adaptation to active change enhances the overall environmental resilience of the enterprise.

[0224] The fused cross-departmental knowledge graph is formed by superimposing real-time business and environmental features on the semantically mapped knowledge graph. This process reflects the semantic enhancement from static to dynamic, from concept to instance. In terms of technical implementation, graph embedding technology is usually used to map real-time feature vectors to new nodes in the knowledge graph, and generate new semantic association edges based on the correlation algorithm between features. For example, the "current capacity utilization" element in the real-time business feature vector can be mapped to the "current capacity utilization" node in the knowledge graph, and establish an association edge with the "output" node. The element value in the real-time feature vector is the attribute of the newly generated node. Through the continuous update of attribute values, the knowledge graph can instantly reflect the latest business status. In the fusion process, attention should be paid to the logical coordination of new nodes, new edges and the original knowledge graph to avoid semantic conflicts. In addition, with the continuous writing of real-time data, the scale of the fused knowledge graph will continue to increase. It is necessary to take optimization measures such as graph partitioning and hierarchical storage to ensure query and reasoning efficiency. In short, the integrated cross-departmental knowledge graph fully combines the domain knowledge that originally focused on conceptual abstraction with real-time business data, giving the knowledge graph the ability to analyze the present and guide the future.

[0225] The integrated cross-departmental knowledge graph becomes an information hub connecting real-time business and the environment, making knowledge-driven collaborative decision-making closer to business reality. By continuously integrating real-time features, the knowledge graph has a more three-dimensional and dynamic grasp of the overall business situation, and the timeliness and pertinence of collaborative optimization have been significantly improved. On the one hand, the integration of real-time business features enables the knowledge graph to accurately depict the strengths and weaknesses of current business operations, explore the "benefit coupling points" of departmental collaboration, and propose targeted optimization measures. On the other hand, after the integration of real-time environmental features, the knowledge graph is more forward-looking in judging the "opportunity-threat" situation of each department. The shift from passive response to active exploration is conducive to the coordinated use of internal and external resources to achieve overtaking on the curve.

[0226] Semantic relevance measures the mutual influence of the businesses of various departments in the integrated knowledge graph. The higher the relevance, the greater the business coupling between departments, and the need for strengthened collaborative management and control. The calculation of semantic relevance is usually based on similarity algorithms of graph data, such as random walk algorithm and node neighbor algorithm. These algorithms take into account the knowledge graph structure (such as the number of shared nodes) and semantics (such as the weight of edges) factors, and can comprehensively evaluate the relevance between departments. In order to improve computing efficiency, a relevance threshold is usually set in the knowledge graph to filter out redundant calculations with extremely weak relevance. The semantic relevance matrix is ​​N×N dimensional, where N is the number of departments in the enterprise. The (i', j') element in the matrix represents the semantic relevance of the i'th department to the j'th department. Due to the asymmetry of the relevance, the matrix is ​​often not a symmetric matrix. The semantic relevance matrix dynamically depicts the intertwined influence network of the businesses of various departments within the enterprise, and has insight into the business dependency relationship of "prosperity and loss together". Relying on the relevance matrix, the weak links of business collaboration can be seen through, providing decision-making references for the systematic promotion of collaborative transformation.

[0227] By dynamically calculating the semantic correlation, the implicit involvement of inter-departmental business is continuously revealed, breaking the "information cocoon" and creating a collaborative atmosphere. From the perspective of correlation, each department is deeply aware of the limitations of "going it alone" and actively seeks collaborative breakthroughs. On the one hand, high correlation indicates great potential for collaborative creation. Departments should focus on opening up key links to achieve complementary advantages and mutual promotion. On the other hand, departments with high negative correlation are "collaborative minefields" and their decision-making errors will harm the overall situation and must be carefully controlled. In addition, the correlation matrix can also be used for departmental performance appraisal, incorporating collaborative performance into the evaluation system, forming a community of interests with shared honor and disgrace, and stimulating collaborative motivation.

[0228] Normalization refers to mapping the original correlation to the interval [0,1] through mathematical transformation to form a unified and intuitive measurement. Common normalization methods include linear normalization, logarithmic normalization, exponential normalization, etc. It is necessary to select an appropriate normalization function according to the distribution of correlation value range. The normalized semantic correlation matrix is ​​convenient for horizontal comparison. Matrix elements close to 1 represent strong correlation between departments, which implies the inherent need of "being in the same boat"; matrix elements close to 0 mean that the business connection between departments is weak, and there is a risk of "going their own way". Through correlation cluster analysis, the departments within the enterprise can be divided into several "collaboration circles". The strong synergy effect within the circle is the "collaboration sample room" that is the focus of breakthrough. However, the cross-circle synergy is weak, and efforts should be made to build a bridge for departmental collaboration. The normalized semantic correlation matrix intuitively presents the interest connection between departments and becomes a "configuration diagram" to guide enterprises to optimize organizational methods and process mechanisms. According to the "collaboration circle" layout of the enterprise structure and the supporting synergy incentive mechanism, the synergy effect can be maximized and the overall competitiveness of the enterprise can be enhanced. By normalizing the correlation matrix, the correlation between departments' interests can be directly depicted, simplifying the difficulty of collaborative management. Traditional department divisions are mostly function-oriented, while the normalized semantic correlation matrix starts from the essence of the business, outlines the implicit departmental interest community, and provides new ideas for corporate organizational reshaping.

[0229] Step S2300 keeps the knowledge graph in sync with business practices by continuously introducing real-time business and environmental data, reducing the time lag caused by using historical data in the knowledge graph construction phase. This "living knowledge graph" enhances the knowledge graph's ability to analyze, trace, and predict trends for business problems by dynamically mapping the business panorama. At the same time, embedding real-time data into a unified semantic system allows business intelligence from different departments to be integrated, associated, and triggered at the semantic level. This provides a data and knowledge foundation for cross-departmental collaborative optimization.

[0230] Step S2400, extract the weights of each node and edge of the fused cross-departmental knowledge graph, dynamically adjust the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix, and generate a cross-departmental collaborative knowledge graph.

[0231] Further, step S2400 includes:

[0232] Step S2410: extract the semantic relevance between departments from the semantic relevance matrix, and for departments whose semantic relevance is higher than a preset relevance threshold θ 3 The first weight adjustment coefficient k 1 To improve the weights of its related nodes and edges, the first weight adjustment coefficient k 1 It is directly proportional to the semantic relevance;

[0233] Step S2420: for the semantic relevance less than or equal to the preset relevance threshold θ 3 The department pairs are adjusted by the second weight coefficient k 2 To reduce the weights of its related nodes and edges, the second weight adjustment coefficient k 2 It is inversely proportional to the semantic relevance.

[0234] Specifically, based on the integrated cross-departmental knowledge graph, considering the closeness of business connections between different departments, the weights of the nodes and edges in the graph are adjusted to form a knowledge graph that can reflect cross-departmental collaborative relationships. Among them, the weight of the node reflects the importance of the node in cross-departmental collaboration, and the weight of the edge reflects the strength of the collaborative relationship between connected nodes. The basis for weight adjustment is the semantic relevance matrix, which quantitatively describes the semantic relevance between different departments based on business data. Weight adjustment includes two directions: increasing weight and reducing weight.

[0235] The semantic relevance matrix quantitatively describes the degree of relevance between different departments at the semantic level of business data and is an important basis for weight adjustment. 3 It is the threshold for judging whether the semantic association between departments is significant. It needs to be set by business experts based on the analysis of the distribution of semantic association, such as 0.8. 3 For example, the semantic correlation between departments E' and F' is 0.9, which means that the two are closely connected in terms of indicator correlation, process connection, and decision-making coordination, and there is a large space for collaborative optimization. Therefore, it is necessary to highlight their status in the collaborative knowledge graph and enhance the weights of related nodes and edges. 1 As the first weight adjustment coefficient, its value increases with the increase of semantic relevance, that is, k 1 It is proportional to the semantic relevance. 1 = semantic relevance, for example, k of department E' and F' 1 Take 0.9. If the original weight of a node between them is 1, the adjusted new weight = 1×(1+k 1 )=1×(1+0.9)=1.9, an increase of 90%. 1 After the adjustment, the connection between highly semantically related departments in the collaborative knowledge graph will be more prominent, and the corresponding indicators, processes, decision-making issues, etc. will become the focus of collaborative management. The increase in weight will tilt resource allocation toward areas with greater collaborative needs and strengthen cross-departmental collaboration.

[0236] For semantic relevance below the threshold θ 3For example, the semantic correlation between departments C' and D' is 0.5, which means that the synergy between the two departments at the business level is relatively weak, and the synergy demand is not the most urgent. In order to focus resources on key synergy areas, it is necessary to reduce the weights of related nodes and edges between C' and D' and weaken their influence in the synergy knowledge graph. 2 As the second weight adjustment coefficient, its value increases as the semantic relevance decreases, that is, k 2 It is inversely proportional to the semantic relevance. 2 = 1-semantic association, for example, the k of departments C' and D' 2 =1-0.5=0.5, if the original weight of a node between them is 1, the adjusted new weight = 1×(1-k 2 )=1×(1-0.5)=0.5, reduced by 50%. 2 After the adjustment, the connection strength of low semantically related departments will be reduced in the collaborative knowledge graph to avoid excessive dispersion of collaborative resources. 2 The application aims to identify areas with relatively weak collaborative associations, while focusing on high semantic association directions and moderately reducing attention and investment in marginal collaborative areas.

[0237] In summary, the weight adjustment in the process of cross-departmental collaborative knowledge graph generation reflects the idea of ​​differentiated collaborative resource allocation based on semantic relevance. High semantic relevance reveals the key areas and focus points of cross-departmental collaboration, which need to be given special attention and resource allocation; while the collaborative needs in low semantic relevance areas are relatively minor, and resource input should be reduced accordingly. This differentiated weight adjustment helps collaborative management focus on key points and improve resource utilization efficiency. The quantified semantic relevance provides an objective basis for weight adjustment, k 1 and k 2 The adaptation of the weight adjustment range and semantic relevance is further achieved, and the balance between strengthening collaboration and focusing on key points is comprehensively considered. It can be foreseen that the cross-departmental collaborative knowledge graph will provide a global perspective and implementation handle for collaborative optimization, and become an important tool for connecting decentralized businesses and stimulating collaborative vitality. Looking to the future, as business complexity continues to increase, cross-departmental collaboration will become the norm, and collaborative knowledge graphs are urgently needed to activate organizational vitality and drive collaborative upgrades.

[0238] Step S3000, generating cross-departmental collaborative decision-making recommendations based on the cross-departmental collaborative knowledge graph.

[0239] Furthermore, step S3000 includes:

[0240] Step S3100, collecting the operation time series data of the nodes in the cross-departmental collaborative knowledge graph, and obtaining the time series dependency features between the nodes based on the operation time series data and the pre-built time series dependency feature model;

[0241] Specifically, the operation records of the process nodes are extracted from the business systems and logs of each department to obtain the time series of each node. Step S3100 characterizes the temporal dependencies between process nodes by analyzing the operation time series generated by business activities. The operation time series data records the start and end time of each business activity. By aligning the operation logs of different departments, a globally unified time series is established.

[0242] In the preprocessing stage, it is necessary to unify the time format and process missing values ​​and outliers to ensure the integrity and accuracy of the time series. Common methods for processing missing values ​​include deleting missing records, filling in the nearest value, interpolation, etc. Outliers can be identified through methods such as box plots and replaced or eliminated.

[0243] The pre-built time series dependency feature model can be a statistical model or a machine learning model, which is used to extract the time series dependency features between nodes from the time series data. Taking the statistical model as an example, the time delay distribution between nodes can be calculated, that is, the time interval distribution from the end time of node G' to the start time of node H'. The concentration of the time delay distribution reflects the strength of the time series dependency between nodes, and the discreteness of the delay distribution reflects the stability of the time series dependency between nodes.

[0244] For example, for two consecutive production process nodes, the time delay distribution is concentrated in 1-2 hours, indicating that there is a close timing dependency between the two processes, the end time of the previous process basically determines the start time of the next process, and the two processes are highly synchronized. If the time delay distribution is discrete, such as some batches are delayed by 1 hour and some are delayed by 5 hours, it indicates that there is a problem with the coordination of the two processes and the timing dependency is unstable.

[0245] Machine learning models such as RNN and LSTM can consider the long-term dependency of time series and explore the deep patterns of the temporal relationship between nodes. The operation time series of a node is predicted by training the model, and the error between the predicted value and the true value is used as a measure of the strength of the temporal dependency. The smaller the prediction error, the greater the influence of other nodes on the operation time of the node, and the stronger the temporal dependency.

[0246] After obtaining the temporal dependency features between nodes, a temporal dependency feature vector can be constructed for each node. For Node , , The timing dependence strengths of are 0.8, 0.5, and 0.1 respectively, then the nodes The timing dependency feature vector of can be expressed as [0.8, 0.5, 0.1]. The timing dependency feature vector describes the dependency relationship of nodes under global timing constraints and is an important input for subsequent fusion analysis.

[0247] Step S3100 mines the implicit timing dependencies between process nodes, which is crucial for optimizing processes and controlling rhythm. Business activities that appear to be parallel may have implicit constraints. Revealing timing dependencies helps to identify bottlenecks in the process and optimize task sequencing and resource allocation. The introduction of pre-built models can efficiently and accurately extract timing features from large-scale time series data. Manual analysis of time series is time-consuming and labor-intensive, and it is difficult to discover complex patterns. By building statistical or machine learning models, timing features can be analyzed automatically and can adapt to different business scenarios. The obtained timing dependency features can quantify the degree of synchronization between nodes and provide early warning indicators for business process monitoring. By tracking changes in the intensity of timing dependencies, abnormal delays or advances of process nodes can be discovered in a timely manner, thereby optimizing rhythm and resource scheduling.

[0248] Step S3200, obtaining semantic interaction features between nodes based on the nodes and edges of the cross-departmental collaborative knowledge graph and the pre-built semantic interaction feature model;

[0249] Specifically, step S3200 learns the semantic interaction patterns between nodes through in-depth exploration of the structured data of the knowledge graph. In the knowledge graph, nodes represent specific business elements (such as indicators, processes, decisions, etc.), and edges represent the associations between business elements. By analyzing the semantic interactions between nodes and edges, key nodes and their impact paths can be discovered.

[0250] The semantic interaction feature describes the intensity and mode of a node's influence on other nodes. A node transmits its influence to adjacent nodes through the edges in the graph, thereby indirectly influencing more distant nodes. The intensity of the influence is related to the shortest path length between nodes and the semantics of the connecting edges.

[0251] After obtaining the node and edge data of the knowledge graph, the first step is to learn the representation of the nodes and edges. Through models such as Skip-gram, CBOW, and DeepWalk, nodes can be mapped to a low-dimensional dense vector space, so that nodes with similar semantics are closer in the vector space. Models such as TransE can be used for representation learning of edges, which can achieve the mapping from the head entity vector to the tail entity vector through translation transformation, encoding the directional semantics of the edge.

[0252] Based on the vector representation of nodes and edges, a pre-trained semantic interaction feature model is used to learn the semantic interaction features of nodes. The semantic interaction feature model uses the Graph Attention Network (GAT). Compared with the traditional graph convolutional network, GAT introduces an attention mechanism to the aggregation of neighbor nodes, so that key neighbors get higher weights, while the influence of non-key neighbors is weakened. This is in line with the actual characteristics of business element interaction.

[0253] The importance of each node in GAT to its neighboring nodes is proportional to the attention weight. The attention weight is obtained by calculating the similarity of the node representation. The higher the similarity, the stronger the semantic association between nodes, and the higher the weight obtained when aggregating the neighboring node information. Each layer of GAT fuses the original features of the node with the information transmitted by the neighboring nodes. After multiple layers of iterative updates, the semantic interaction feature representation of a node is finally obtained.

[0254] By analyzing the semantic interaction characteristics between nodes, we can discover the association patterns of business elements. For example, when the quality management indicators are abnormal, the nodes with high semantic interaction intensity with the quality indicators may be upstream factors such as supplier changes and equipment failures, revealing the root cause of the quality problem. Downstream nodes with high semantic interaction intensity with quality indicators may include customer satisfaction, return rate, etc., indicating that the impact of quality problems has spread. An impact network diagram is drawn around the semantic interaction characteristics of the quality abnormality nodes, which is helpful for traceability analysis and decision making.

[0255] The graph attention network can adaptively focus on the importance of different neighbor nodes and extract key features of semantic interactions. Business elements are complex and not all associations are valid. The attention mechanism gives higher weights to key nodes and key paths, so that decisions can be focused on the main contradictions. The multi-layer graph attention network can explore high-order semantic associations between nodes and reveal the implicit dependencies of business elements. Directly related nodes can be found in the shallow layers of the graph, while indirectly related nodes require more layers of information transmission to be mined. This helps to systematically understand business problems and formulate fundamental solutions. Semantic interaction features can be used to predict the consequences of node state changes and assist in risk assessment. Understanding the impact transmission path between nodes can predict the abnormal scope of a business indicator, take measures in advance, and achieve precise governance.

[0256] Step S3300, integrate the temporal dependency features and semantic interaction features, combine with logical rules for reasoning, and generate cross-departmental collaborative decision-making recommendations.

[0257] Further, step S3300 includes:

[0258] Step S3310, fusing the temporal dependency feature and the semantic interaction feature to form a fused feature vector;

[0259] Step S3320, constructing a logic rule library and establishing a logic rule reasoning engine;

[0260] Step S3330, input the fused feature vector into the logic rule reasoning engine, perform reasoning, and generate cross-departmental collaborative decision-making recommendations.

[0261] Specifically, feature fusion is the process of integrating node features extracted from two different perspectives into a unified representation. The temporal dependency feature describes the association pattern of a node with other nodes in the time dimension, such as the time-lag correlation and Granger causality of node pairs. The semantic interaction feature describes the importance of a node in the graph semantic network and its semantic similarity with other nodes. These two types of features describe the role of a node in the knowledge graph from the perspectives of temporal dependency and semantic association, respectively. Fusion of them into a vector can form a more comprehensive feature description of the node.

[0262] Common methods of feature fusion include feature concatenation and feature weighted summation. Feature concatenation is to connect two feature vectors end to end to form a longer vector. The advantage of feature vector concatenation is that it retains the complete information of the original features, but the disadvantage is that the dimension of the fused features is higher. When the original feature dimension is already very high, concatenation will further aggravate the dimensionality disaster. Another fusion method is feature weighted summation. Through weighted summation, features from different sources can be mapped to the same dimension to form a semantic "mixture". The selection of weight coefficients can be based on the discriminative ability of the features, giving higher weights to features that have a greater impact on the prediction results. The fusion feature dimension generated by the weighted summation of features is the same as the original feature dimension, which alleviates the dimensionality explosion problem to a certain extent. However, some original feature information may be lost during the fusion process.

[0263] The fused feature vector expresses the joint semantics of temporal dependency features and semantic interaction features, and can better support subsequent tasks. Through feature fusion, heterogeneous features can be compared in the same semantic space to discover the intrinsic connection between cross-departmental business data. This feature fusion is a common technique in data mining and machine learning, and can be applied to clustering, classification, association rule mining, etc. In this scenario, by extracting and fusing different side features of knowledge graph nodes, conditions are created for mining cross-departmental collaborative relationships. The fused features can be input into the prediction model to find out the indicators and processes that are more likely to have cross-influences, identify the key objects that need cross-departmental coordination, and discover the feature patterns that promote collaboration or lead to dis-cooperation. Based on this, a handle for communication and collaboration between departments and an entry point for joint optimization can be formed.

[0264] Logical rules are IF-THEN formal statements used to characterize the causal dependency between variables. Logical rule reasoning matches the rule antecedent conditions, triggers the rules when the conditions are met, and draws corresponding conclusions. Building a logical rule base is to summarize the business domain knowledge and formalize it as a series of rules. These rules can be derived from the experience of domain experts, business constraints, management specifications, etc., and include causal relationship judgments in production, quality, supply, etc. For example:

[0265] IF [defect rate increases] AND [raw material quality decreases] THEN [incoming material quality inspection fails];

[0266] IF [key process capability index>1.33] THEN [the process meets production requirements];

[0267] IF [raw material prices rise] AND [order volume decreases] THEN [reduce raw material purchases];

[0268] In the rule, the antecedent (after IF and before THEN) represents the triggering condition, and the consequent (after THEN) represents the inference result. Converting business knowledge into logical rules allows it to be organized and applied in a computer-understandable way.

[0269] On this basis, a logical rule inference engine needs to be established to realize automatic triggering and matching of rules. The inference engine matches the input data with the antecedents of the rules in the rule base. When a rule is found to be satisfied, it triggers the rule and outputs the inference result defined by the consequent. The implementation of the inference engine is divided into three steps: matching, conflict resolution and execution. Matching is to check whether the input data can meet the antecedent conditions of one or more rules. When multiple rule antecedents are satisfied, conflicts may arise. It is necessary to use a conflict resolution strategy to decide which rule to execute, such as based on rule priority or based on rule specificity (rules with more specific conditions take precedence). After the rule is determined, it is executed and its inference result is returned.

[0270] For example, input data shows: node The features of meet the requirements of "quality index decreases" and "production speed increases", and match the rule "IF [quality index decreases] AND [production speed increases] THEN [quality control problems may exist]". The rule is triggered and executed, and the inference result "quality control problems may exist" is output.

[0271] The advantage of logical rule reasoning is that it is explainable, and the reasoning results can be traced back to the triggered rules. This makes the reasoning process transparent to business personnel and easy to interpret. At the same time, the rule base can be continuously updated and expanded to add new business knowledge and improve the adaptability of the system. Its disadvantage is that it requires domain experts to participate in knowledge acquisition and rule construction, which is costly. In addition, the coverage of the rule base is limited by the breadth of expert knowledge, and the generalization ability to deal with unknown scenarios is not strong. It needs to be combined with other machine learning methods to make up for its limitations.

[0272] Node fusion features can be used as input for logical rule reasoning. The reasoning engine identifies the feature combination that meets the preset business logic by matching node features with rule antecedents, and then obtains the corresponding reasoning results. This process can be applied to abnormal node identification, bottleneck mining, opportunity discovery, etc.

[0273] Taking the identification of abnormal nodes as an example, the following rules are designed:

[0274] IF [quality index decrease > 10%] AND [process parameter change] THEN [process change leads to quality abnormality];

[0275] The quality index of a certain node dropped by 15%, and its process parameters changed. This feature combination hits the above rule. Based on this, the inference engine determines that "process change leads to quality abnormality" at this node. The system generates an early warning message and pushes it to the quality and production departments, prompting them to pay attention to this node, analyze the impact of the process change, and take corrective measures.

[0276] Intelligent decision-making assistance based on logical rule reasoning has three positive effects:

[0277] 1. Trigger-based decision-making. Passive waiting is no longer suitable for the dynamically changing business environment. Rule triggering based on real-time data streams can achieve a closed decision loop, turning management from passive response to active prevention.

[0278] 2. Collaborative decision-making. Departmental walls and information islands limit the decision-making perspective. Cross-departmental data integration and rule reasoning help clarify decision-making dependencies and consolidate concerted actions.

[0279] 3. Traceable decision-making. The lack of an effective mechanism to trace the cause of a decision is not conducive to continuous optimization. Rule reasoning can provide a basis for decision-making, making the decision-making process explainable and traceable, and providing support for review learning.

[0280] For example, a company uses this solution to optimize cross-departmental production collaboration. By integrating production, quality, and supply data to build a knowledge graph, the timing and semantic features of key nodes are extracted and a rule base is built. The system detects that the characteristics of a material supply node meet the rule "IF [supplier performance assessment is less than 80 points] AND [material quality fluctuates greatly] AND [delivery cycle is extended] THEN [change alternative suppliers]". In order to prevent the spread of supply risks to the production link, the inference engine promptly outputs the decision suggestion of "changing alternative suppliers" and analyzes the production task nodes that may be affected based on the knowledge graph. The production management department adjusts the production plan accordingly, the quality department focuses on tracking related batches of products, and multiple departments form a coordinated response.

[0281] This case demonstrates the typical path of integrating features and rule reasoning to support cross-departmental collaborative decision-making, namely, opening up data flows, integrating business rules, and connecting departmental actions. This solution effectively improves the timeliness, effectiveness, and explainability of cross-departmental collaboration, helping companies strengthen digital operations and improve risk response capabilities. In the future, we can combine the correction effect data to dynamically optimize feature extraction and rule setting, so that the collaborative mechanism can continue to iterate and evolve.

[0282] Example 2

[0283] This embodiment provides a reasoning system based on semantic association and logical rules of knowledge graph on the basis of embodiment 1, such as Figure 8 As shown, including:

[0284] Graph construction module: used to obtain a first data set across functional departments, construct a structured data table based on the first data set, and construct a cross-departmental knowledge graph based on the structured data table;

[0285] Graph optimization module: extract indicator nodes from the cross-departmental knowledge graph, identify semantic conflict indicators, and generate an indicator semantic conflict list; based on the indicator semantic conflict list, construct a semantically mapped cross-departmental knowledge graph; obtain a second data set across functional departments, and based on the second data set and the semantically mapped cross-departmental knowledge graph, obtain a fused cross-departmental knowledge graph and a semantic relevance matrix; extract the weights of each node and edge of the fused cross-departmental knowledge graph, and dynamically adjust the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix to generate a cross-departmental collaborative knowledge graph;

[0286] Decision reasoning module: Generate cross-departmental collaborative decision-making recommendations based on the cross-departmental collaborative knowledge graph.

[0287] In the graph construction module, the first data set includes indicator definition documents, business process description documents and historical decision records across functional departments; the cross-functional departments include the production department, quality management department and supply chain department.

[0288] In the graph construction module, constructing a structured data table according to the first data set includes:

[0289] Step S1210, extracting the indicator name A, the first relationship type and the indicator name B from the indicator definition document to form an indicator information triple; wherein the indicator name A is the subject of the indicator information triple, the first relationship type is the predicate of the indicator information triple, and the indicator name B is the object of the indicator information triple; the first relationship type includes a calculation relationship, an influence relationship and a composition relationship;

[0290] Step S1220, extracting the process node A', the second relationship type and the process node B' from the business process description document to form a process triple; wherein the process node A' is the subject of the process triple, the second relationship type is the predicate of the process triple, and the process node B' is the object of the process triple; the second relationship type includes a temporal relationship and a logical relationship;

[0291] Step S1230, extracting decision events, third relationship types, decision problems or solutions from historical decision records to form decision information triples; wherein the decision event is the subject of the decision information triple, the third relationship type is the predicate of the decision information triple, and the decision problem or solution is the object of the decision information triple; the third relationship type includes the relationship between the decision event and the decision problem, and the relationship between the decision event and the decision solution;

[0292] Step S1240, constructing a structured data table based on the indicator information triples, the process triples and the decision information triples, the structured data table including triple subjects, triple predicates and triple objects.

[0293] In the graph construction module, the construction of a cross-departmental knowledge graph based on a structured data table includes:

[0294] Step S1310, mapping triple subjects and triple objects into nodes in a cross-departmental knowledge graph, and mapping triple predicates into directed edges between nodes; the nodes in the cross-departmental knowledge graph include indicator nodes, process nodes, and decision nodes;

[0295] Step S1320, extracting attribute information from the first data set, adding attributes to the nodes according to the attribute information, and generating node attributes; the node attributes include indicator node attribute information, process node attribute information, and decision node attribute information; the indicator node attribute information includes indicator name, indicator definition, and responsible department;

[0296] Step S1330, annotating the nodes in the cross-departmental knowledge graph with semantic types and adding initial node weights to the nodes;

[0297] Step S1340, annotating the semantic types of the directed edges in the cross-departmental knowledge graph and adding initial edge weights to the directed edges;

[0298] Step S1350, optimizing the cross-departmental knowledge graph.

[0299] The step S1350 includes:

[0300] Step S1351, calculate the semantic similarity SI between nodes in the cross-department knowledge graph, and set the semantic similarity SI greater than the preset first similarity threshold θ 1 The nodes of are defined as synonymous concept nodes;

[0301] Step S1352, adding synonymous relationship edges between synonymous concept nodes;

[0302] Step S1353, identify the upper and lower concept nodes in the cross-departmental knowledge graph, and add belonging relationship edges between the upper and lower concept nodes.

[0303] In the graph optimization module, the generated indicator semantic conflict list includes:

[0304] Step S2110, retrieving all indicator nodes in the cross-departmental knowledge graph and extracting indicator node attribute information;

[0305] Step S2120, vectorizing the indicator name and indicator definition in the indicator node attribute information to form an indicator vector of the indicator corresponding to the indicator node;

[0306] Step S2130, calculate the similarity S between the indicator vectors of each indicator 1 , set the second similarity threshold θ 2 , for the similarity S 1 Higher than θ 2 However, the indicator pairs with different responsible departments are marked as semantically conflicting indicators and added to the indicator semantic conflict list.

[0307] In the graph optimization module, the cross-departmental knowledge graph after constructing the semantic mapping includes:

[0308] Step S2210, for each indicator pair in the indicator semantic conflict list, query a preset semantic mapping rule library to match the mapping rule; the semantic mapping rule is expressed in the form of IF-THEN;

[0309] Step S2220, according to the matched mapping rule, adding a mapping relationship edge and a mapping attribute between two semantically conflicting indicators in the indicator pair to form a cross-departmental semantic mapping layer;

[0310] Step S2230, integrating the cross-departmental semantic mapping layer into the cross-departmental knowledge graph to form a cross-departmental knowledge graph after semantic mapping.

[0311] In the graph optimization module, the fused cross-departmental knowledge graph and semantic relevance matrix include:

[0312] Step S2310, forming a real-time business feature vector according to the real-time business parameters across functional departments;

[0313] Step S2320, forming a real-time environment feature vector based on the real-time environment data across functional departments;

[0314] Step S2330, fusing the real-time business feature vector and the real-time environment feature vector in the semantically mapped cross-departmental knowledge graph to form a fused cross-departmental knowledge graph;

[0315] Step S2340, based on the fused cross-department knowledge graph, calculate the semantic relevance between departments and generate an N×N semantic relevance matrix, where N is the number of departments;

[0316] Step S2350, normalizing the semantic relevance matrix.

[0317] In the graph optimization module, generating a cross-departmental collaborative knowledge graph includes:

[0318] Step S2410: extract the semantic relevance between departments from the semantic relevance matrix, and for departments whose semantic relevance is higher than a preset relevance threshold θ 3 The first weight adjustment coefficient k 1 To improve the weights of its related nodes and edges, the first weight adjustment coefficient k 1 It is directly proportional to the semantic relevance;

[0319] Step S2420: for the semantic relevance less than or equal to the preset relevance threshold θ 3 The department pairs are adjusted by the second weight coefficient k 2 To reduce the weights of its related nodes and edges, the second weight adjustment coefficient k 2 It is inversely proportional to the semantic relevance.

[0320] In the decision-making reasoning module, the generation of cross-departmental collaborative decision suggestions based on the cross-departmental collaborative knowledge graph includes:

[0321] Step S3100, collecting the operation time series data of the nodes in the cross-departmental collaborative knowledge graph, and obtaining the time series dependency features between the nodes based on the operation time series data and the pre-built time series dependency feature model;

[0322] Step S3200, obtaining semantic interaction features between nodes based on the nodes and edges of the cross-departmental collaborative knowledge graph and the pre-built semantic interaction feature model;

[0323] Step S3300, integrate the temporal dependency features and semantic interaction features, combine with logical rules for reasoning, and generate cross-departmental collaborative decision-making recommendations.

[0324] The step S3300 includes:

[0325] Step S3310, fusing the temporal dependency feature and the semantic interaction feature to form a fused feature vector;

[0326] Step S3320, constructing a logic rule library and establishing a logic rule reasoning engine;

[0327] Step S3330, input the fused feature vector into the logic rule reasoning engine, perform reasoning, and generate cross-departmental collaborative decision-making recommendations.

[0328] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0329] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A reasoning method based on semantic association and logical rules of knowledge graph, characterized in that: The method comprises: Obtain a first data set across functional departments, construct a structured data table based on the first data set, and construct a cross-departmental knowledge graph based on the structured data table; Extract indicator nodes from the cross-departmental knowledge graph, identify semantic conflict indicators, and generate an indicator semantic conflict list; construct a semantically mapped cross-departmental knowledge graph based on the indicator semantic conflict list; obtain a second data set across functional departments, and obtain a fused cross-departmental knowledge graph and a semantic relevance matrix based on the second data set and the semantically mapped cross-departmental knowledge graph; extract the weights of each node and edge of the fused cross-departmental knowledge graph, and dynamically adjust the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix to generate a cross-departmental collaborative knowledge graph; Based on the cross-departmental collaborative knowledge graph, cross-departmental collaborative decision-making recommendations are generated.

2. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 1, characterized in that: The first data set includes indicator definition documents, business process description documents and historical decision records of cross-functional departments; the cross-functional departments include production departments, quality management departments and supply chain departments; The step of constructing a structured data table according to the first data set includes: Extracting the indicator name A, the first relationship type and the indicator name B from the indicator definition document to form an indicator information triple; wherein the indicator name A is the subject of the indicator information triple, the first relationship type is the predicate of the indicator information triple, and the indicator name B is the object of the indicator information triple; Extracting process node A', the second relationship type and process node B' from the business process description document to form a process triple; wherein process node A' is the subject of the process triple, the second relationship type is the predicate of the process triple, and process node B' is the object of the process triple; Extracting decision events, third relation types, decision problems or solutions from historical decision records to form decision information triples; wherein the decision events are the subjects of the decision information triples, the third relation types are the predicates of the decision information triples, and the decision problems or solutions are the objects of the decision information triples; Based on the indicator information triples, process triples and decision information triples, a structured data table is constructed. The structured data table includes triple subjects, triple predicates and triple objects.

3. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 2 is characterized in that: The construction of a cross-departmental knowledge graph based on a structured data table includes: Mapping triple subjects and triple objects into nodes in a cross-departmental knowledge graph, and mapping triple predicates into directed edges between nodes; the nodes in the cross-departmental knowledge graph include indicator nodes, process nodes, and decision nodes; Extracting attribute information from the first data set, adding attributes to the nodes according to the attribute information, and generating node attributes; the node attributes include indicator node attribute information, process node attribute information, and decision node attribute information; the indicator node attribute information includes indicator name, indicator definition, and responsible department; Label the nodes in the cross-departmental knowledge graph with semantic types and add initial node weights to the nodes; Label the semantic types of directed edges in the cross-departmental knowledge graph and add initial edge weights to the directed edges; optimize the cross-departmental knowledge graph.

4. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 3 is characterized in that: The optimization of the cross-departmental knowledge graph includes: Calculate the semantic similarity SI between nodes in the cross-departmental knowledge graph, and define the nodes whose semantic similarity SI is greater than the preset first similarity threshold θ1 as synonymous concept nodes; Add synonymous relationship edges between synonymous concept nodes; Identify the upper and lower concept nodes in the cross-departmental knowledge graph, and add belonging relationship edges between the upper and lower concept nodes.

5. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 3 is characterized in that: The generated indicator semantic conflict list includes: Retrieve all indicator nodes in the cross-departmental knowledge graph and extract indicator node attribute information; Vectorize the indicator name and indicator definition in the indicator node attribute information to form an indicator vector of the indicator corresponding to the indicator node; Calculate the similarity S1 between the indicator vectors of each indicator, set the second similarity threshold θ2, and for indicator pairs with similarity S1 higher than θ2 but different responsible departments, mark them as semantic conflict indicators and add them to the indicator semantic conflict list.

6. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 5, characterized in that: The cross-departmental knowledge graph constructed after semantic mapping includes: For each indicator pair in the indicator semantic conflict list, query the preset semantic mapping rule library to match the mapping rule; the semantic mapping rule is expressed in the form of IF-THEN; According to the matched mapping rules, the mapping relationship edge and mapping attributes between the two semantic conflicting indicators in the indicator pair are added to form a cross-departmental semantic mapping layer; Integrate the cross-departmental semantic mapping layer into the cross-departmental knowledge graph to form a cross-departmental knowledge graph after semantic mapping.

7. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 6 is characterized in that: The second data set includes real-time business parameters and real-time environmental data across functional departments; The obtained integrated cross-departmental knowledge graph and semantic relevance matrix include: Based on the real-time business parameters across functional departments, a real-time business feature vector is formed; Based on the real-time environmental data across functional departments, a real-time environmental feature vector is formed; In the cross-departmental knowledge graph after semantic mapping, the real-time business feature vector and the real-time environment feature vector are integrated to form a fused cross-departmental knowledge graph; Based on the integrated cross-departmental knowledge graph, the semantic relevance between departments is calculated to generate an N×N semantic relevance matrix, where N is the number of departments.

8. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 7 is characterized in that: The generating of cross-departmental collaborative knowledge graph includes: Extracting the semantic relevance between departments from the semantic relevance matrix, for department pairs whose semantic relevance is higher than a preset relevance threshold θ3, using a first weight adjustment coefficient k1 to increase the weights of their related nodes and edges, where the first weight adjustment coefficient k1 is proportional to the semantic relevance; For department pairs whose semantic relevance is less than or equal to the preset relevance threshold θ3, the weights of their related nodes and edges are reduced by the second weight adjustment coefficient k2, and the second weight adjustment coefficient k2 is inversely proportional to the semantic relevance.

9. The reasoning method based on semantic association and logical rules of knowledge graph according to claim 8, characterized in that: The generation of cross-departmental collaborative decision-making suggestions based on the cross-departmental collaborative knowledge graph includes: Collect the operation time series data of nodes in the cross-departmental collaborative knowledge graph, and obtain the time series dependency features between nodes based on the operation time series data and the pre-built time series dependency feature model; According to the nodes and edges of the cross-departmental collaborative knowledge graph and the pre-built semantic interaction feature model, the semantic interaction features between nodes are obtained; Fuse the temporal dependency features and the semantic interaction features to form a fused feature vector; Build a logic rule library and establish a logic rule reasoning engine; The fused feature vector is input into the logic rule reasoning engine for reasoning to generate cross-departmental collaborative decision-making recommendations.

10. A reasoning system based on semantic associations and logical rules of knowledge graphs, which is used to implement a reasoning method based on semantic associations and logical rules of knowledge graphs as described in any one of claims 1 to 9, characterized in that: The system comprises: Graph construction module: used to obtain a first data set across functional departments, construct a structured data table based on the first data set, and construct a cross-departmental knowledge graph based on the structured data table; Graph optimization module: extract indicator nodes from the cross-departmental knowledge graph, identify semantic conflict indicators, and generate an indicator semantic conflict list; based on the indicator semantic conflict list, construct a semantically mapped cross-departmental knowledge graph; obtain a second data set across functional departments, and based on the second data set and the semantically mapped cross-departmental knowledge graph, obtain a fused cross-departmental knowledge graph and a semantic relevance matrix; extract the weights of each node and edge of the fused cross-departmental knowledge graph, and dynamically adjust the weights of each node and edge of the fused cross-departmental knowledge graph based on the semantic relevance matrix to generate a cross-departmental collaborative knowledge graph; Decision reasoning module: Generate cross-departmental collaborative decision-making recommendations based on the cross-departmental collaborative knowledge graph.

Images