CN-122021837-A - Enterprise information collaborative management method and system based on multi-source data fusion

CN122021837ACN 122021837 ACN122021837 ACN 122021837ACN-122021837-A

Abstract

The application discloses an enterprise information collaborative management method and system based on multi-source data fusion, relating to the technical field of collaborative management, the method comprises the steps of carrying out full-mode serialization and context reconstruction on heterogeneous data streams containing orders and work orders, and converting discrete forms into serialized contexts with semantic association. And then, extracting the business rules as system-level constraint injection, and guiding the large language model to mine the structured relation of the cross-domain data under the rule framework. And then, carrying out logic verification and deterministic alignment on the reasoning result by utilizing a double gating threshold value, and eliminating semantic ambiguity. Based on the above, supernode fusion and attribute aggregation are performed on the initial map, and a collaborative knowledge map for mapping global business logic is constructed. And finally, analyzing long-chain dependence and generating a collaborative instruction through map traversal and sub-graph pattern matching. Therefore, the accuracy of entity alignment in a complex environment can be effectively improved, and the intellectualization and automation of enterprise information collaborative management are realized.

Inventors

XU BAOWEI
LI SHUJU
XU BAOPENG

Assignees

浙江九烁光电工程技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (9)

1. The enterprise information collaborative management method based on multi-source data fusion is characterized by comprising the following steps of: s1, carrying out full-mode serialization and context reconstruction on a structured table stream containing order and work order information to obtain a serialized context corpus; s2, extracting business rule logic based on metadata constraint information, and injecting the business rule logic as system-level constraint into a serialization context corpus to obtain a mode enhanced reasoning prompt set; S3, inputting the mode enhanced reasoning prompt set into a pre-trained large language model to obtain a structured reasoning result set; S4, carrying out logic consistency check and deterministic alignment on the structured reasoning result set by utilizing a double gating threshold value to obtain a check entity alignment pair; S5, performing supernode fusion and attribute aggregation operation on the pre-constructed initial map according to the alignment of the check entity so as to obtain a global collaborative knowledge map; and S6, traversing the global collaborative knowledge graph to perform sub-graph pattern matching on the acquired specific business event so as to obtain a collaborative execution instruction set.
2. The enterprise information collaborative management method based on multi-source data fusion according to claim 1, wherein step S1 comprises: carrying out natural language template mapping of the structured data on each record in the structured table stream to obtain a structured description text set; performing entity preservation denoising processing on the unstructured text stream to obtain a pure unstructured text set; mapping the structured description text set and the pure unstructured text set to corresponding space-time anchor points for physical splicing to obtain a serialization context corpus.
3. The enterprise information collaborative management method based on multi-source data fusion according to claim 1, wherein step S2 comprises: performing natural language transcoding of hard business rules on formalized logic in the metadata constraint information to obtain a natural language constraint set; Carrying out sliding window extraction and pairing on the serialization context corpus and the natural language constraint set to obtain a prompt unit list to be assembled; and based on a preset template structure, performing dynamic character stitching on the prompt unit list to be assembled to obtain a mode enhanced reasoning prompt set.
4. The enterprise information collaborative management method based on multi-source data fusion according to claim 1, characterized in that step S3 comprises: Inputting the mode enhanced reasoning prompt set into a pre-trained large language model to obtain an original reasoning text set containing a reasoning process and a final conclusion; Performing regular extraction and segmentation of the logic components on the original reasoning text set to obtain an extracted logic component list; And performing JSON serialization and numerical type standardization on the extracted logical component list to obtain a structured reasoning result set.
5. The enterprise information collaborative management method based on multi-source data fusion according to claim 1, wherein step S4 comprises: Performing double-gating-based candidate screening on the structured reasoning result set to obtain a high-confidence candidate pair list and an exact negative case auxiliary set; Mapping the high confidence candidate pair list into an undirected graph structure, and searching a connected subgraph by utilizing a graph traversal algorithm to obtain a temporary logical connected component set; and combining the exact negative example auxiliary set to perform triangle logic paradox detection and conflict fusing on the temporary logic connected component set so as to output alignment of the verification entity.
6. The collaborative management method for enterprise information based on multi-source data fusion according to claim 5, wherein performing double-gating-based candidate screening on a structured inference result set to obtain a high-confidence candidate pair list and an exact negative case auxiliary set, comprises: Introducing a Bayesian risk perception confidence calculation model, carrying out nonlinear calibration on the original large language model confidence in the structured reasoning result set to obtain a Bayesian risk perception score, calculating a dynamic safety threshold value based on the source authority degree, and carrying out risk-based screening decision by comparing the Bayesian risk perception score with the dynamic safety threshold value; the Bayes risk perception score is calculated based on the confidence coefficient of the original large language model, the source authority degree function and the evidence intensity factor, wherein the evidence intensity factor is calculated based on the reasoning path length of the thinking chain; the dynamic safety threshold is calculated based on a preset reference threshold and a risk penalty coefficient; And only when the calculated Bayesian risk perception score is higher than the dynamic safety threshold, the corresponding reasoning result is included in the high-confidence candidate pair list, otherwise, the high-confidence candidate pair list is classified into an exact negative example auxiliary set.
7. The enterprise information collaborative management method based on multi-source data fusion according to claim 1, characterized in that step S5 comprises: performing supernode instantiation and ID remapping on the check entity alignment pair to obtain a node merging mapping table; Performing conflict resolution and aggregation on original attributes pointing to the same supernode through a node merging mapping table based on a source credibility priority strategy so as to obtain a merged attribute vector set; And based on the node merging mapping table and the merging attribute vector set, executing node replacement and redirecting the associated edges in the graph database so as to obtain the global collaborative knowledge graph.
8. The enterprise information collaborative management method based on multi-source data fusion according to claim 1, characterized in that step S6 comprises: carrying out sub-graph isomorphism search on a specific business event on a global collaborative knowledge graph so as to identify a matched event sub-graph set meeting specific topology and attribute conditions; Performing business logic calculation on the real-time attributes of the nodes in the matched event sub-graph set based on the decision strategy table to obtain an abstract action target list; The abstract action target list is converted into machine codes adapting to the target system interface protocol and the authentication mode through a protocol adapter to obtain a collaborative execution instruction set.
9. An enterprise information collaborative management system based on multi-source data fusion is characterized by comprising: The serialization and context reconstruction module is used for carrying out full-mode serialization and context reconstruction on the heterogeneous data of the structured table stream containing the order and the work order information so as to obtain a serialized context corpus; the mode enhancement module is used for extracting business rule logic based on metadata constraint information, and injecting the business rule logic serving as system-level constraint into the serialization context corpus to obtain a mode enhancement reasoning prompt set; the large language model reasoning module is used for inputting the mode enhanced reasoning prompt set into the pre-trained large language model to obtain a structured reasoning result set; The double gating verification module is used for carrying out logic consistency verification and deterministic alignment on the structured reasoning result set by utilizing a double gating threshold value so as to obtain a verification entity alignment pair; The knowledge spectrum fusion module is used for executing supernode fusion and attribute aggregation operation on the pre-constructed initial spectrum according to alignment of the check entity so as to obtain a global collaborative knowledge spectrum; and the event matching and instruction generation module is used for traversing the global collaborative knowledge graph to perform sub-graph pattern matching on the acquired specific business event so as to obtain a collaborative execution instruction set.

Description

Enterprise information collaborative management method and system based on multi-source data fusion Technical Field The application relates to the technical field of collaborative management, in particular to an enterprise information collaborative management method and system based on multi-source data fusion. Background Along with the vigorous development of global digital economy, enterprise informatization construction has been stepped into deep water areas, ERP, MES, CRM and other heterogeneous information systems are widely applied to production and operation activities of various enterprises. The systems generate massive order streams, work order streams, logistics records and interaction logs all weather in the running process, and form data assets of an enterprise core. In order to cope with increasingly strong market competition, the agile scheduling of production resources and the fine management and control of business processes are realized, the data island formed by the system architecture difference and business barriers is broken, and the construction of an information collaborative management system based on multi-source data fusion is particularly urgent. Enterprises are required to observe business dynamics from a global perspective by carrying out deep fusion and full-link association analysis on multi-source heterogeneous data so as to improve cross-department collaboration efficiency and decision response speed, thereby maintaining competitive advantages in a complex market environment. However, although the existing data integration technology alleviates the information splitting problem to a certain extent, the technology still faces a serious technical bottleneck when facing high-dimensional, complex and dynamically changing cross-domain service data. Traditional enterprise information collaboration schemes rely mostly on predefined static ETL rules or shallow database schema mapping techniques, which lack the ability to perceive and understand deep semantic contexts of data. In practical applications, order information and work order information are often scattered in different service domains, and data modalities are different, so that semantic mapping among cross-system entities is often ambiguous. In the prior art, in an industrial data environment filled with noise, heterogeneous entities with complex corresponding relations are difficult to accurately identify and dynamically align, and the problem of semantic disambiguation in cross-domain data fusion cannot be effectively solved. In addition, in a complex business scenario involving long-link dependence, due to insufficient construction precision of a topological structure of a bottom layer entity relationship, a system often cannot accurately capture logic association between events, so that when a business abnormality is faced, it is difficult to penetrate layer-by-layer data representation to accurately position a root cause. The limitation caused by low entity alignment precision and fuzzy topological relation severely restricts the intelligent level and the automatic execution efficiency of the enterprise information collaborative management. Therefore, there is a need for an optimized enterprise information collaborative management method and system. Disclosure of Invention The present application has been made to solve the above-mentioned technical problems. According to one aspect of the application, there is provided an enterprise information collaborative management method based on multi-source data fusion, comprising: s1, carrying out full-mode serialization and context reconstruction on a structured table stream containing order and work order information to obtain a serialized context corpus; s2, extracting business rule logic based on metadata constraint information, and injecting the business rule logic as system-level constraint into a serialization context corpus to obtain a mode enhanced reasoning prompt set; S3, inputting the mode enhanced reasoning prompt set into a pre-trained large language model to obtain a structured reasoning result set; S4, carrying out logic consistency check and deterministic alignment on the structured reasoning result set by utilizing a double gating threshold value to obtain a check entity alignment pair; S5, performing supernode fusion and attribute aggregation operation on the pre-constructed initial map according to the alignment of the check entity so as to obtain a global collaborative knowledge map; and S6, traversing the global collaborative knowledge graph to perform sub-graph pattern matching on the acquired specific business event so as to obtain a collaborative execution instruction set. According to another aspect of the present application, there is provided an enterprise information collaborative management system based on multi-source data fusion, including: The serialization and context reconstruction module is used for carrying out full-mode serialization and c