Enterprise operation data analysis method and system based on knowledge graph

Abstract

The invention discloses an enterprise operation data analysis method and system based on a knowledge graph, and relates to the technical field of data processing, and the method comprises the steps: collecting an enterprise multi-source heterogeneous operation data set, and carrying out the standardization processing and multi-dimensional entity extraction; performing knowledge graph modeling and dynamic updating optimization based on the enterprise operation multi-dimensional entity set, and constructing an enterprise operation knowledge graph; constructing an operation data analysis task list and performing prediction driving on the enterprise operation knowledge graph; and obtaining enterprise operation period data, performing risk prediction reasoning and operation decision analysis based on the operation data analysis task engine, determining a target enterprise operation decision scheme, and performing operation decision control through the target enterprise operation decision scheme. The technical problems of enterprise operation data islanding, weak knowledge analysis and reasoning ability and decision lag in the prior art are solved, and the technical effects of improving operation data analysis and reasoning intelligence and providing prospective risk prediction and accurate operation decision support are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an enterprise operation data analysis method and system based on a knowledge graph. BACKGROUND

[0002] A large amount of multi-source heterogeneous data is generated in the daily operation process of an enterprise, covering multiple dimensions such as finance, production, supply chain, market, and human resources. How to efficiently integrate, analyze, and utilize these data to support enterprise decision-making, optimize operation processes, and predict potential risks has become a key to modern enterprise operation management. Traditional enterprise operation data analysis is difficult to cope with complex and changing business environments, resulting in lagging, one-sided, and lack of explainability of analysis results. A knowledge graph can effectively integrate multi-source heterogeneous data and provide semantic association and context understanding capabilities by modeling a graph structure of entities, relationships, and attributes. It can convert scattered operation data into an interconnected knowledge network, support dynamic updating and complex reasoning. However, when constructing an enterprise operation knowledge graph, the standardization processing of multi-source heterogeneous data is inefficient, entity extraction is inefficient, dynamic maintenance and optimization of the knowledge graph are insufficient, and the adaptability of analysis tasks and business scenarios is poor, lacking reasoning for risk prediction and decision support.

[0003] Therefore, in the related art at the present stage, there are technical problems of enterprise operation data islanding, weak knowledge analysis reasoning ability, and decision lag. SUMMARY

[0004] The present application provides an enterprise operation data analysis method and system based on a knowledge graph, which solves the technical problems of enterprise operation data islanding, weak knowledge analysis reasoning ability, and decision lag in the prior art, and achieves the technical effects of improving the intelligence of operation data analysis reasoning and providing forward-looking risk prediction and precise operation decision support.

[0005] The present application provides an enterprise operation data analysis method based on a knowledge graph, which comprises: collecting a set of multi-source heterogeneous operation data of an enterprise, standardizing and processing the set of multi-source heterogeneous operation data of the enterprise, and extracting multi-dimensional entities to obtain a set of multi-dimensional entities of enterprise operation; based on the set of multi-dimensional entities of enterprise operation, modeling and dynamically updating and optimizing a knowledge graph to construct an enterprise operation knowledge graph; constructing a list of operation data analysis tasks, predicting and driving the enterprise operation knowledge graph according to the list of operation data analysis tasks to generate an operation data analysis task engine; obtaining data during enterprise operation, performing risk prediction reasoning and operation decision analysis on the data during enterprise operation based on the operation data analysis task engine, determining a target enterprise operation decision scheme, and performing operation decision control through the target enterprise operation decision scheme.

[0006] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: structurally classifying the enterprise multi-source heterogeneous operation data set to obtain structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data; performing a preprocessing step analysis on the structured enterprise operation data, the semi-structured enterprise operation data, and the unstructured enterprise operation data respectively to construct a heterogeneous data preprocessing program; performing standardized preprocessing on the structured enterprise operation data, the semi-structured enterprise operation data, and the unstructured enterprise operation data respectively by using the heterogeneous data preprocessing program to obtain a usable enterprise operation data set; and performing multi-dimensional entity extraction on the usable enterprise operation data set to obtain an enterprise operation multi-dimensional entity set.

[0007] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: constructing a three-dimensional entity system according to enterprise operation business, the three-dimensional entity system including a business field dimension, a user role dimension, and a product operation dimension; performing entity type association on dimension information in the three-dimensional entity system to obtain a three-dimensional associated operation entity type label; performing entity content subdivision on the three-dimensional associated operation entity type label respectively to determine a three-dimensional entity type label associated content set; and performing multi-dimensional entity extraction on the usable enterprise operation data set according to the three-dimensional entity type label associated content set to obtain the enterprise operation multi-dimensional entity set.

[0008] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: performing associated level division on the enterprise operation multi-dimensional entity set to obtain enterprise operation entity associated level information; performing cascade analysis on the enterprise operation multi-dimensional entity set according to the enterprise operation entity associated level information to construct an operation knowledge graph architecture; performing content filling and logical association on the enterprise operation multi-dimensional entity set based on the enterprise multi-source heterogeneous operation data set to obtain an operation entity associated node content set; mapping the operation entity associated node content set to the operation knowledge graph architecture to perform knowledge graph modeling and dynamic update optimization, and obtaining an enterprise operation knowledge graph.

[0009] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: filling the enterprise operation multi-dimensional entity set based on the available enterprise operation data set to obtain an operation multi-dimensional entity content set; mining business rules from the available enterprise operation data set to obtain an enterprise operation business rule set; performing business logic analysis on the operation multi-dimensional entity content set based on the enterprise operation business rule set to obtain an entity content logical relationship set; and logically associating each entity content in the operation multi-dimensional entity content set as an entity content node set according to the entity content logical relationship set to obtain the operation entity associated node content set.

[0010] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: mapping the operation entity associated node content set to the operation knowledge graph architecture for knowledge graph modeling to construct an initial operation knowledge graph; monitoring enterprise operation update data streams in real time according to a preset period; performing entity extraction association and dynamic update optimization on the initial operation knowledge graph based on the enterprise operation update data streams to obtain the enterprise operation knowledge graph.

[0011] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: performing entity extraction association on the initial operation knowledge graph based on the enterprise operation update data streams to obtain an operation update entity content set; performing conflict detection correction on the operation update entity content set using the initial operation knowledge graph to obtain available operation entity data; and performing dynamic update optimization on the initial operation knowledge graph based on the available operation entity data to obtain the enterprise operation knowledge graph.

[0012] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: sequentially associating each analysis task in the operation data analysis task list with the enterprise operation knowledge graph to obtain an operation analysis task associated knowledge node set; performing historical operation data mining based on the operation analysis task associated knowledge node set to obtain an operation analysis task associated knowledge data set; respectively performing prediction training optimization on the operation analysis task associated knowledge data set to construct an operation data analysis task model set; and coupling and driving the operation data analysis task model set to generate the operation data analysis task engine.

[0013] In a possible implementation, the enterprise operation data analysis method based on the knowledge graph further performs the following processing: according to the characteristic information of the operation analysis task associated knowledge data set, selecting an operation analysis task model structure set; using the operation analysis task model structure set to respectively perform identification prediction training and cross-validation optimization on the operation analysis task associated knowledge data set, and constructing the operation data analysis task model set.

[0014] The application also provides an enterprise operation data analysis system based on a knowledge graph, which comprises: a data acquisition module, configured to acquire an enterprise multi-source heterogeneous operation data set, perform standardization processing and multi-dimensional entity extraction on the enterprise multi-source heterogeneous operation data set, and obtain an enterprise operation multi-dimensional entity set; a knowledge graph construction module, configured to perform knowledge graph modeling and dynamic update optimization based on the enterprise operation multi-dimensional entity set, and construct an enterprise operation knowledge graph; a prediction driving module, configured to construct an operation data analysis task list, perform prediction driving on the enterprise operation knowledge graph according to the operation data analysis task list, and generate an operation data analysis task engine; and an operation decision analysis module, configured to acquire enterprise operation period data, perform risk prediction reasoning and operation decision analysis on the enterprise operation period data based on the operation data analysis task engine, determine a target enterprise operation decision scheme, and perform operation decision control through the target enterprise operation decision scheme.

[0015] The enterprise operation data analysis method and system based on the knowledge graph provided in the application acquire an enterprise multi-source heterogeneous operation data set and perform standardization processing and multi-dimensional entity extraction; perform knowledge graph modeling and dynamic update optimization based on an enterprise operation multi-dimensional entity set, and construct an enterprise operation knowledge graph; construct an operation data analysis task list and perform prediction driving on the enterprise operation knowledge graph; acquire enterprise operation period data, perform risk prediction reasoning and operation decision analysis based on an operation data analysis task engine, determine a target enterprise operation decision scheme, and perform operation decision control through the target enterprise operation decision scheme. The technical problems of enterprise operation data islanding, weak knowledge analysis reasoning capability and decision lag in the prior art are solved, and the technical effects of improving operation data analysis reasoning intelligence and providing forward-looking risk prediction and precise operation decision support are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. In the present application, a flowchart is used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, according to the needs, various steps can be processed in reverse order or simultaneously. Meanwhile, other operations can be added to these processes, or a step or several steps can be removed from these processes.

[0017] Figure 1 A knowledge graph-based enterprise operation data analysis method flowchart is provided for the embodiments of the present application.

[0018] Figure 2 A knowledge graph-based enterprise operation data analysis system structure diagram is provided for the embodiments of the present application.

[0019] Legend: data acquisition module 10, knowledge graph construction module 20, prediction driving module 30, operation decision analysis module 40. DETAILED DESCRIPTION

[0020] The above description is only a summary of the technical solutions of the present application. In order to make the technical means of the present application more clear, the embodiments of the present application can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following specific embodiments of the present application are described.

[0021] In order to make the purposes, technical solutions and advantages of the present application more clear, the following will combine the drawings to make a further detailed description of the present application. The described embodiments should not be regarded as limiting the present application, and all other embodiments obtained by those skilled in the art without making creative efforts belong to the scope of protection of the present application.

[0022] In the following description, "some embodiments" are involved, which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subset of all possible embodiments, and can be combined with each other without conflict. The term "first\second" is only to distinguish similar objects, and does not represent the specific order of the objects. The terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art in the technical field of the present application. The terms used herein are only for the purpose of describing the embodiments of the present application.

[0023] The embodiments of the present application provide a knowledge graph-based enterprise operation data analysis method, as shown in Figure 1 The method comprises the following steps: Step S100, acquiring enterprise multi-source heterogeneous operation data set, standardizing and multi-dimensional entity extracting the enterprise multi-source heterogeneous operation data set to obtain enterprise operation multi-dimensional entity set.

[0024] Preferably, the enterprise multi-source heterogeneous operation data set is collected from enterprise resource planning systems, customer relationship management systems, supply chain management systems, production execution systems, financial software, Internet of Things logs, market reports, etc., and can include structured data such as SQL tables in relational databases, semi-structured data such as JSON, XML format API interface data, log files, and unstructured data such as Word documents, PDF contracts, email texts, and text in pictures. Then, the enterprise multi-source heterogeneous operation data set is standardized, including data cleaning and format standardization, and multi-dimensional entity extraction is performed using named entity recognition in natural language processing to identify and extract entity objects representing key businesses, such as customer A, product B, order C, supplier D, equipment E, and post F, from the standardized enterprise operation data, and finally obtain an enterprise operation multi-dimensional entity set, and the enterprise operation multi-dimensional entity contains entity types and entity attributes.

[0025] Further, step S100 further includes step S110 of classifying the enterprise multi-source heterogeneous operation data set into structured data, semi-structured data, and unstructured data; step S120 of respectively preprocessing the structured data, semi-structured data, and unstructured data to construct a heterogeneous data preprocessing program; step S130 of using the heterogeneous data preprocessing program to respectively standardize the structured data, semi-structured data, and unstructured data to obtain a usable enterprise operation data set; and step S140 of extracting multi-dimensional entities from the usable enterprise operation data set to obtain an enterprise operation multi-dimensional entity set.

[0026] Preferably, the enterprise multi-source heterogeneous operation data set is classified into structured data, semi-structured data, and unstructured data, and the structured data, semi-structured data, and unstructured data are determined, wherein the structured data refers to data with strict and fixed patterns, usually stored in two-dimensional table form, such as business tables in relational databases, each row of which is a data record and each column is a defined field; the semi-structured data refers to data with no fixed patterns, usually defined by labels or tags, such as JSON, XML format API interface return data, log files, and emails; and the unstructured data refers to data without predefined data patterns, such as Word documents, PDF contracts, PPT presentations, pictures, videos, audio files, and their text contents.

[0027] Preferably, the preprocessing step is performed on structured enterprise operation data, semi-structured enterprise operation data and unstructured enterprise operation data respectively, that is, data cleaning and conversion rules are written, specifically, for structured enterprise operation data, data deduplication, handling null values, format conversion, outlier correction and the like are performed, for semi-structured enterprise operation data, JSON / XML parsing, label extraction, log pattern matching and the like are performed, for unstructured enterprise operation data, optical character recognition is used to extract text from pictures or PDFs, natural language processing is used for text segmentation, stop word removal, entity recognition and the like, and finally the cleaning and conversion rules of the three types of data are coded into heterogeneous data preprocessing programs, which can automatically determine the type of input data and call the corresponding processing unit.

[0028] Preferably, the heterogeneous data preprocessing program is used to standardize the preprocessing of structured enterprise operation data, semi-structured enterprise operation data and unstructured enterprise operation data, eliminate inconsistencies and errors between data, and finally output a set of usable enterprise operation data with uniform data mode and format and no missing, error or duplicate values; using natural language processing, entities representing real-world objects are identified and extracted from the set of usable enterprise operation data, and attributes and type information of different business dimensions are defined as labels, for example, a customer entity has a credit rating attribute in the financial dimension, a purchase history attribute in the sales dimension, and a support ticket attribute in the service dimension, and finally a set of enterprise operation multi-dimensional entities is obtained.

[0029] Further, step S140 further includes step S141 of constructing a three-dimensional entity system according to enterprise operation business, wherein the three-dimensional entity system includes a business domain dimension, a user role dimension and a product operation dimension; step S142 of performing entity type association on the dimension information in the three-dimensional entity system to obtain three-dimensional associated operation entity type labels; step S143 of performing entity content subdivision on the three-dimensional associated operation entity type labels respectively to determine a three-dimensional entity type label associated content set; and step S144 of performing multi-dimensional entity extraction on the set of usable enterprise operation data according to the three-dimensional entity type label associated content set to obtain the set of enterprise operation multi-dimensional entities.

[0030] Preferably, a three-dimensional entity system for analyzing and classifying enterprise entities is built according to enterprise operation business, including business field dimension, user role dimension and product operation dimension, wherein the business field dimension refers to the core functions of the enterprise, such as supply chain, production and manufacturing, marketing, sales, human resources, financial management, customer service, etc., to ensure coverage of all key activity fields of the enterprise; the user role dimension refers to various internal and external roles interacting with the enterprise, such as internal employees, customers, suppliers, partners, system administrators, etc., to ensure identification of participating entity objects; the product operation dimension refers to the products / services provided by the enterprise and their life cycle, such as products, service packages, orders, projects, work orders, logistics documents, etc., to ensure identification of entities around the core value provided by the enterprise.

[0031] Preferably, the dimension information in the three-dimensional entity system is associated with entity types, i.e., the business field dimension, user role dimension and product operation dimension are combined and associated to generate a comprehensive and non-redundant entity type list, similar to creating a three-dimensional matrix, each intersection point defining an entity type, and finally obtaining three-dimensional associated operation entity type labels, each three-dimensional associated operation entity type label defining a unique entity type, such as business field sales, user role customer, product operation order, corresponding to the entity type label customer order.

[0032] Preferably, the three-dimensional associated operation entity type labels are respectively subdivided into entity content, i.e., defining entity patterns or data structures to clearly define the entities and entity attributes contained in the entity type labels, and then determining the three-dimensional entity type label associated content set; finally, the available enterprise operation data set is subjected to multi-dimensional entity extraction according to the three-dimensional entity type label associated content set, i.e., according to each entity type and its attributes, performing corresponding SQL queries on the database of the available enterprise operation data set or parsing the corresponding fields from the file, and filling the data into the predetermined entity structure, and finally obtaining the enterprise operation multi-dimensional entity set, each entity having the labels of business field dimension, user role dimension and product operation dimension.

[0033] Step S200, knowledge graph modeling and dynamic updating optimization based on the enterprise operation multi-dimensional entity set, to build an enterprise operation knowledge graph.

[0034] The step S200 further comprises the following steps: S210, performing association level division on the enterprise operation multi-dimensional entity set to obtain enterprise operation entity association level information; S220, performing cascade analysis on the enterprise operation multi-dimensional entity set according to the enterprise operation entity association level information to construct an operation knowledge graph architecture; S230, performing content filling and logical association on the enterprise operation multi-dimensional entity set based on the enterprise multi-source heterogeneous operation data set to obtain an operation entity association node content set; and S240, mapping the operation entity association node content set to the operation knowledge graph architecture to perform knowledge graph modeling and dynamic update optimization, thereby obtaining an enterprise operation knowledge graph.

[0035] Preferably, the enterprise operation multi-dimensional entity set is divided into association levels, that is, all entity types in the enterprise operation multi-dimensional entity set are subjected to semantic relationship analysis and hierarchical classification. Specifically, logical relationship types between different entity types, such as belonging, purchasing, producing, and being responsible for, are analyzed to determine the superior-inferior or subordinate relationship between entities, and a type hierarchy is constructed. For example, entities are the top layer, financial entities and production entities are subclasses thereof, and financial entities can be further divided into customers and suppliers. Then, enterprise operation entity association level information is output, including a semantic mode defining the relationship types and type hierarchy between entity types.

[0036] Preferably, the enterprise operation multi-dimensional entity set is subjected to cascade analysis according to the enterprise operation entity association level information, that is, logical analysis and deduction are performed to determine the chain effect and path between the association level information. For example, an order is associated with a customer, and the customer belongs to a region. Through cascade analysis, the query path from the order to the region can be analyzed, and an operation knowledge graph architecture is constructed, including the storage structure, indexing method, and partitioning strategy of the knowledge graph. Then, the enterprise operation multi-dimensional entity set is subjected to content filling based on the enterprise multi-source heterogeneous operation data set, that is, each entity is filled with specific data attributes, and logical association is performed, that is, relationship links are established between entities according to business rules. For example, according to the foreign key relationship in the order table, a “belongs to” relationship between the order entity and the customer entity is constructed. Then, an operation entity association node content set is output, containing all entity instances, their attributes, and relationship instances.

[0037] Preferably, finally, the operation entity associated node content set is mapped to the operation knowledge graph architecture through an ETL tool or a custom script for knowledge graph modeling, the operation entity associated node content is organized into nodes and edges according to a predetermined operation knowledge graph architecture and semantic mode, and an initial operation knowledge graph is constructed; a mechanism for continuous operation is established to dynamically update and optimize the initial operation knowledge graph, specifically, data changes of a business system, such as new orders and updated customer information, are monitored, and nodes and edges in the graph are automatically updated through incremental updating or timed batch processing, while the graph is indexed and optimized, the storage structure is adjusted, and the like, according to query performance, and finally an enterprise operation knowledge graph containing real-time business data and complex relationships and being capable of being efficiently queried and reasoned is output.

[0038] Further, step S230 further includes step S231 of filling the operation multi-dimensional entity set with content based on the available enterprise operation data set to obtain an operation multi-dimensional entity content set; step S232 of mining business rules from the available enterprise operation data set to obtain an enterprise operation business rule set; step S233 of analyzing business logic of the operation multi-dimensional entity content set based on the enterprise operation business rule set to obtain an entity content logical relationship set; and step S234 of logically associating each entity content in the operation multi-dimensional entity content set as an entity content node set according to the entity content logical relationship set to obtain the operation entity associated node content set.

[0039] Preferably, specific operation data in the available enterprise operation data set is mapped and filled into the entity attribute of the enterprise operation multi-dimensional entity set, for example, for the customer entity type, its customer name attribute is mapped to the name field, and its credit rating attribute is mapped to the credit rating field, thereby outputting the operation multi-dimensional entity content set containing all specific entity instances and attribute values; business rule mining is performed on the available enterprise operation data set, that is, entity association rules are extracted and induced from the business processes, data constraints, industry knowledge and historical data of the enterprise by analyzing the data mode, thereby outputting the enterprise operation business rule set for establishing the entity relationship of the enterprise operation data. Then, based on the enterprise operation business rule set, business logic analysis is performed on the operation multi-dimensional entity content set to infer the specific relationship between the entity instances, for example, a record exists in the order table, then the order belongs to the corresponding customer of the customer table, all order entity instances are traversed, and the relationship type existing between all entity nodes is output, thereby outputting the entity content logical relationship set; finally, each entity content in the operation multi-dimensional entity content set is logically associated as an entity content node set, that is, an association link is created between the entity instances to form a complete and interconnected data structure, that is, the edges of the graph are constructed, each edge has a starting node, a terminal node and a relationship type, and finally the operation entity association node content set is output, containing the node set of all entity instances with specific attribute values and the edge set of all relationship instances with types connecting these nodes.

[0040] Further, step S240 further comprises step S241 of mapping the operation entity association node content set to the operation knowledge graph architecture for knowledge graph modeling to construct an initial operation knowledge graph; step S242 of real-time monitoring of enterprise operation update data flow according to a preset period; and step S243 of performing entity extraction association and dynamic update optimization on the initial operation knowledge graph based on the enterprise operation update data flow to obtain the enterprise operation knowledge graph.

[0041] Preferably, the operation entity association node content set containing all node and edge data is imported into the operation knowledge graph architecture in batches through an ETL script or a data import tool for knowledge graph modeling. The operation entity association node content is created as a graph structure, i.e., the entity becomes a node with a label and an attribute, and the relationship becomes an edge with a type, thereby generating an initial operation knowledge graph. A preset period defines the listening frequency, such as every minute or every few hours. Then, the enterprise operation update data stream is listened to in real time according to the preset period, i.e., the incremental data of the enterprise operation is obtained by listening to the database change log, API interface call, or scanning the incremental file, including the data records that have been newly added, modified, or deleted since the last update. Then, the initial operation knowledge graph is associated with entity extraction according to the enterprise operation update data stream, including performing entity recognition and relationship establishment on the incremental data, and performing dynamic update optimization, including incremental insertion, attribute update, invalid node deletion, and creating an index for new hot data and adjusting the storage structure, etc., to maintain the query efficiency of the graph and keep it synchronized with the real business state. Finally, a live enterprise operation knowledge graph that can continuously evolve and reflect the latest operation state of the enterprise in real time is output.

[0042] Further, step S243 further includes step S2431 of performing entity extraction association on the initial operation knowledge graph based on the enterprise operation update data stream to obtain an operation update entity content set; step S2432 of performing conflict detection and correction on the operation update entity content set using the initial operation knowledge graph to obtain available operation entity data; and step S2433 of performing dynamic update optimization on the initial operation knowledge graph based on the available operation entity data to obtain the enterprise operation knowledge graph.

[0043] Preferably, the initial operation knowledge graph is associated with entity extraction based on the enterprise operation update data stream. The entity objects and the relationships between entities are identified from the enterprise operation update data stream as nodes and edges, respectively, and then the operation update entity content set is output, containing node and edge information to be added, modified, or deleted, such as new sales data being extracted as an order node and a relationship. Then, the operation update entity content set is subjected to conflict detection and correction, including comparing the operation update entity content set with the initial operation knowledge graph to check for existence conflicts, consistency conflicts, and authority conflicts, and processing and correcting the detected conflicts according to predefined business rules to obtain available operation entity data. The initial operation knowledge graph is dynamically updated and optimized, i.e., node and edge addition, modification, and deletion operations are performed, and performance optimization operations are executed, such as creating an index for a large number of newly added similar nodes to speed up the query. Finally, an enterprise operation knowledge graph reflecting the latest enterprise operation state and controlled data quality is output.

[0044] Step S300, constructing an operation data analysis task list, and performing prediction driving on the enterprise operation knowledge graph according to the operation data analysis task list to generate an operation data analysis task engine.

[0045] Step S300 further comprises step S310, sequentially associating each analysis task in the operation data analysis task list with the enterprise operation knowledge graph to obtain an operation analysis task associated knowledge node set; step S320, performing historical operation data mining based on the operation analysis task associated knowledge node set to obtain an operation analysis task associated knowledge data set; step S330, respectively performing prediction training optimization on the operation analysis task associated knowledge data set to construct an operation data analysis task model set; and step S340, coupling and driving the operation data analysis task model set to generate the operation data analysis task engine.

[0046] Preferably, the operation data analysis task list is constructed according to the core target and business of the enterprise, and contains a plurality of specific business problems or targets, such as predicting high-end customer churn risk, analyzing potential impact of supply chain disruption, identifying cross-selling opportunities, etc. Then, the enterprise operation knowledge graph is driven for prediction according to the operation data analysis task list. Specifically, the abstract analysis tasks in the operation data analysis task list are parsed and mapped to the entity nodes of the enterprise operation knowledge graph, including analyzing the entity nodes corresponding to each task based on semantic matching, such as determining all nodes labeled as customers and having the attribute of VIP in the enterprise operation knowledge graph and the associated nodes of the nodes such as orders, service work orders, customer feedback, etc. for the task of predicting high-end customer churn risk, and then outputting the operation analysis task associated knowledge node set.

[0047] Preferably, based on the operation analysis task associated knowledge node set, historical operation data mining is performed, that is, from the knowledge graph and the related historical data storage, the node attributes and their historical states and relationship evolution are extracted as feature data and label data to form training samples. For example, for the task of predicting the risk of loss of high-end customers, the current and historical characteristics of the customers are extracted and labeled, and then the operation analysis task associated knowledge data set is output, which includes various attribute features and corresponding labels extracted from historical data, to form a structured machine learning data set. According to the type of the task, different machine learning models are selected, such as gradient boosting tree, graph neural network, and logistic regression, and the operation analysis task associated knowledge data set is used for prediction training and optimization. Through cross-validation and hyperparameter tuning, the model performance is optimized, so that the model can discover patterns from the enterprise operation knowledge graph and make predictions, and then obtain the operation data analysis task model set, such as the customer loss prediction model and the supply chain risk identification model. Finally, the operation data analysis task model set is coupled and driven, that is, the various prediction models included in the operation data analysis task model set are integrated to obtain an operation data analysis task engine, which can automatically call the corresponding prediction model according to the received analysis task and extract the latest feature data of the required nodes from the latest enterprise operation knowledge graph in real time.

[0048] Further, step S330 further includes step S331 of selecting an operation analysis task model structure set according to the characteristic information of the operation analysis task associated knowledge data set; and step S332 of performing identification prediction training and cross-validation optimization on the operation analysis task associated knowledge data set using the operation analysis task model structure set to construct the operation data analysis task model set.

[0049] Preferably, according to the task type, data size, feature type, and label condition of the operation analysis task associated knowledge data set, the most likely machine learning model is selected, such as logistic regression, random forest, and gradient boosting machine, to form an operation analysis task model structure set; and then the operation analysis task model structure set is used to perform identification prediction training and cross-validation optimization on the operation analysis task associated knowledge data set. Specifically, the features in the operation analysis task associated knowledge data set are used as input, and the predetermined labels are used as target to perform prediction training on the operation analysis task model structure set, learn the mapping relationship between them, obtain the corresponding prediction model, and then divide the training data into multiple subsets of similar size for cross-validation of the prediction model. Through hyperparameter tuning, the model is optimized to achieve the best performance, and finally the operation data analysis task model corresponding to multiple analysis tasks is obtained to form the operation data analysis task model set.

[0050] In step S400, the enterprise operation period data is acquired, the risk prediction reasoning and operation decision analysis are performed on the enterprise operation period data based on the operation data analysis task engine, the target enterprise operation decision scheme is determined, and the operation decision control is performed through the target enterprise operation decision scheme.

[0051] Preferably, the business data is collected in real time from each business system of the enterprise through an API interface, a database log capture, or the like, the enterprise operation period data is obtained, the enterprise operation period data is input into the operation data analysis task engine, each preset analysis task is executed, the risk prediction reasoning and operation decision analysis are performed, specifically, the operation data analysis task engine first associates and fuses the enterprise operation period data with the enterprise operation knowledge graph, calculates real-time features for corresponding entities based on the latest data of the enterprise operation knowledge graph, such as the total order amount of the customer in the last 7 days, then calls a prediction model of a corresponding analysis task, such as a customer loss risk model, performs prediction according to the calculated real-time features, outputs a risk probability value as a prediction result, simultaneously performs operation decision analysis according to an embedded business rule, determines a structured decision scheme list, and selects an optimal scheme as the target enterprise operation decision scheme from the decision scheme list, finally performs operation decision control through the target enterprise operation decision scheme, for example, the target enterprise operation decision scheme is pushed to a corresponding business personnel workbench for review and execution, or the target enterprise operation decision scheme is directly called through an API to automatically execute a business system interface, finally, the forward-looking risk prediction and precise operation decision support are provided, the active prevention of operation risks and the precise decision of business opportunities are realized.

[0052] In the foregoing, the enterprise operation data analysis method based on the knowledge graph according to the embodiments of the present application is described in detail. Next, the enterprise operation data analysis system based on the knowledge graph according to the embodiments of the present application will be described with reference to the accompanying drawings. Figure 1 The enterprise operation data analysis method based on the knowledge graph according to the embodiments of the present application is described in detail. Next, the enterprise operation data analysis system based on the knowledge graph according to the embodiments of the present application will be described with reference to the accompanying drawings. Figure 2 The enterprise operation data analysis system based on the knowledge graph according to the embodiments of the present application is described in detail. Next, the enterprise operation data analysis system based on the knowledge graph according to the embodiments of the present application will be described with reference to the accompanying drawings.

[0053] The enterprise operation data analysis system based on the knowledge graph according to the embodiments of the present application is described in detail. Next, the enterprise operation data analysis system based on the knowledge graph according to the embodiments of the present application will be described with reference to the accompanying drawings. Figure 2 As shown in FIG. 1, the enterprise operation data analysis system based on the knowledge graph includes a data acquisition module 10, a knowledge graph construction module 20, a prediction driving module 30, and an operation decision analysis module 40.

[0054] The data collection module 10 is configured to collect enterprise multi-source heterogeneous operation data sets, perform standardization processing and multi-dimensional entity extraction on the enterprise multi-source heterogeneous operation data sets, and obtain enterprise operation multi-dimensional entity sets. The knowledge graph construction module 20 is configured to perform knowledge graph modeling and dynamic update optimization based on the enterprise operation multi-dimensional entity sets, and construct an enterprise operation knowledge graph. The prediction driving module 30 is configured to construct an operation data analysis task list, perform prediction driving on the enterprise operation knowledge graph according to the operation data analysis task list, and generate an operation data analysis task engine. The operation decision analysis module 40 is configured to obtain enterprise operation period data, perform risk prediction reasoning and operation decision analysis on the enterprise operation period data based on the operation data analysis task engine, determine a target enterprise operation decision scheme, and perform operation decision control through the target enterprise operation decision scheme.

[0055] In the following, the specific configuration of the data collection module 10 will be described in detail. The data collection module 10 further comprises: structuring and classifying the enterprise multi-source heterogeneous operation data sets to obtain structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data; performing a preprocessing step analysis on the structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data respectively to construct a heterogeneous data preprocessing program; performing standardization preprocessing on the structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data respectively using the heterogeneous data preprocessing program to obtain a usable enterprise operation data set; and performing multi-dimensional entity extraction on the usable enterprise operation data set to obtain the enterprise operation multi-dimensional entity set.

[0056] In the following, the specific configuration of the data collection module 10 will be described in detail. The data collection module 10 further comprises: constructing a three-dimensional entity system according to enterprise operation business, the three-dimensional entity system comprising a business domain dimension, a user role dimension, and a product operation dimension; performing entity type association on the dimension information in the three-dimensional entity system to obtain three-dimensional associated operation entity type labels; performing entity content subdivision on the three-dimensional associated operation entity type labels respectively to determine a three-dimensional entity type label associated content set; and performing multi-dimensional entity extraction on the usable enterprise operation data set according to the three-dimensional entity type label associated content set to obtain the enterprise operation multi-dimensional entity set.

[0057] In the following, the specific configuration of the knowledge graph construction module 20 will be described in detail. The knowledge graph construction module 20 further comprises: performing hierarchical division on the enterprise operation multi-dimensional entity set to obtain enterprise operation entity hierarchical information; performing cascade analysis on the enterprise operation multi-dimensional entity set according to the enterprise operation entity hierarchical information to construct an operation knowledge graph architecture; filling content and logically associating the enterprise operation multi-dimensional entity set based on the enterprise multi-source heterogeneous operation data set to obtain an operation entity associated node content set; mapping the operation entity associated node content set to the operation knowledge graph architecture for knowledge graph modeling and dynamic update optimization to obtain an enterprise operation knowledge graph.

[0058] In the following, the specific configuration of the knowledge graph construction module 20 will be described in detail. The knowledge graph construction module 20 further comprises: filling content of the enterprise operation multi-dimensional entity set based on the available enterprise operation data set to obtain an operation multi-dimensional entity content set; mining business rules from the available enterprise operation data set to obtain an enterprise operation business rule set; performing business logic analysis on the operation multi-dimensional entity content set based on the enterprise operation business rule set to obtain an entity content logical relationship set; logically associating each entity content in the operation multi-dimensional entity content set as an entity content node set according to the entity content logical relationship set to obtain the operation entity associated node content set.

[0059] In the following, the specific configuration of the knowledge graph construction module 20 will be described in detail. The knowledge graph construction module 20 further comprises: mapping the operation entity associated node content set to the operation knowledge graph architecture for knowledge graph modeling to construct an initial operation knowledge graph; real-time monitoring of enterprise operation update data streams according to a preset period; performing entity extraction association and dynamic update optimization on the initial operation knowledge graph based on the enterprise operation update data stream to obtain the enterprise operation knowledge graph.

[0060] In the following, the specific configuration of the knowledge graph construction module 20 will be described in detail. The knowledge graph construction module 20 further comprises: performing entity extraction association on the initial operation knowledge graph based on the enterprise operation update data stream to obtain an operation update entity content set; performing conflict detection correction on the operation update entity content set using the initial operation knowledge graph to obtain available operation entity data; performing dynamic update optimization on the initial operation knowledge graph based on the available operation entity data to obtain the enterprise operation knowledge graph.

[0061] Below, the specific configuration of the prediction driving module 30 will be described in detail. The prediction driving module 30 further comprises: sequentially associating each analysis task in the operation data analysis task list with the enterprise operation knowledge graph to obtain an operation analysis task associated knowledge node set; performing historical operation data mining based on the operation analysis task associated knowledge node set to obtain an operation analysis task associated knowledge data set; respectively performing prediction training optimization on the operation analysis task associated knowledge data set to construct an operation data analysis task model set; and coupling driving the operation data analysis task model set to generate the operation data analysis task engine.

[0062] Below, the specific configuration of the prediction driving module 30 will be described in detail. The prediction driving module 30 further comprises: selecting an operation analysis task model structure set according to the characteristic information of the operation analysis task associated knowledge data set; and respectively performing identification prediction training and cross-validation optimization on the operation analysis task associated knowledge data set using the operation analysis task model structure set to construct the operation data analysis task model set.

[0063] The enterprise operation data analysis system based on the knowledge graph provided in the embodiments of the present application can execute the enterprise operation data analysis method based on the knowledge graph provided in any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0064] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, and each unit and module included is only divided according to the functional logic, but is not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for the convenience of mutual differentiation, and do not limit the protection scope of the present application.

[0065] The above specific embodiments do not constitute a limitation on the protection scope of the present application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A knowledge graph-based enterprise operation data analysis method, characterized in that, The method includes: Collect enterprise multi-source heterogeneous operation datasets, and perform standardization processing and multi-dimensional entity extraction on the enterprise multi-source heterogeneous operation datasets to obtain enterprise operation multi-dimensional entity sets. Based on the aforementioned multi-dimensional entity set of enterprise operations, knowledge graph modeling and dynamic updating optimization are performed to construct an enterprise operations knowledge graph. Construct an operational data analysis task list, and use the enterprise operational knowledge graph to generate an operational data analysis task engine based on the operational data analysis task list; The system acquires data during the enterprise's operation period, performs risk prediction and reasoning and operational decision analysis on the data based on the operational data analysis task engine, determines the target enterprise's operational decision plan, and conducts operational decision control through the target enterprise's operational decision plan.

2. The knowledge graph-based enterprise operation data analysis method as described in claim 1, characterized in that, The obtained multidimensional entity set of enterprise operations includes: The multi-source heterogeneous enterprise operation dataset is classified in a structured manner to obtain structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data. The structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data are respectively analyzed through preprocessing steps to construct a heterogeneous data preprocessing program; The heterogeneous data preprocessing program is used to standardize and preprocess the structured enterprise operation data, semi-structured enterprise operation data, and unstructured enterprise operation data respectively to obtain a usable enterprise operation dataset. Multidimensional entity extraction is performed on the available enterprise operation dataset to obtain the enterprise operation multidimensional entity set.

3. The knowledge graph-based enterprise operation data analysis method as described in claim 2, characterized in that, The step of extracting multidimensional entities from the available enterprise operation dataset to obtain an enterprise operation multidimensional entity set includes: Based on the company's operational business, a three-dimensional entity system is constructed, which includes the business domain dimension, the user role dimension, and the product operation dimension. Entity type association is performed on the information of each dimension in the three-dimensional entity system to obtain three-dimensional associated operational entity type labels; The entity content of the three-dimensional associated operational entity type tags is further subdivided to determine the set of associated content for the three-dimensional entity type tags. The available enterprise operation dataset is subjected to multidimensional entity extraction based on the content set associated with the three-dimensional entity type labels to obtain the enterprise operation multidimensional entity set.

4. The knowledge graph-based enterprise operation data analysis method as described in claim 2, characterized in that, The construction of the enterprise operation knowledge graph includes: The enterprise operation multidimensional entity set is hierarchically divided to obtain enterprise operation entity hierarchical information; Based on the hierarchical information of the enterprise's operational entities, a cascade analysis is performed on the multidimensional entity set of the enterprise's operations to construct an operational knowledge graph architecture. Based on the enterprise's multi-source heterogeneous operation dataset, the enterprise's multi-dimensional operation entity set is populated with content and logically associated to obtain the operation entity associated node content set. The content set of associated nodes of the operational entity is mapped to the operational knowledge graph architecture for knowledge graph modeling and dynamic updating and optimization to obtain the enterprise operational knowledge graph.

5. The knowledge graph-based enterprise operation data analysis method as described in claim 4, characterized in that, The obtained content set of associated nodes of the operating entity includes: Based on the available enterprise operation dataset, the enterprise operation multidimensional entity set is classified and populated to obtain the operation multidimensional entity content set; Business rule mining is performed on the available enterprise operation dataset to obtain a set of enterprise operation business rules; Based on the enterprise operation business rule set, business logic analysis is performed on the operation multidimensional entity content set to obtain the entity content logical relationship set; According to the logical relationship set of entity content, each entity content in the operational multidimensional entity content set is logically associated as an entity content node set to obtain the operational entity associated node content set.

6. The knowledge graph-based enterprise operation data analysis method as described in claim 4, characterized in that, The acquisition of the enterprise operation knowledge graph includes: The content set of associated nodes of the operational entity is mapped to the operational knowledge graph architecture for knowledge graph modeling, and an initial operational knowledge graph is constructed. Monitor enterprise operational data updates in real time according to a preset cycle; Based on the enterprise operation update data stream, entity extraction, association, and dynamic update optimization are performed on the initial operation knowledge graph to obtain the enterprise operation knowledge graph.

7. The knowledge graph-based enterprise operation data analysis method as described in claim 6, characterized in that, The process of obtaining the enterprise operation knowledge graph includes: Based on the enterprise operation update data stream, entity extraction and association are performed on the initial operation knowledge graph to obtain the operation update entity content set; The initial operational knowledge graph is used to perform conflict detection and correction on the operational update entity content set to obtain usable operational entity data. The initial operational knowledge graph is dynamically updated and optimized based on the available operational entity data to obtain the enterprise operational knowledge graph.

8. The knowledge graph-based enterprise operation data analysis method as described in claim 1, characterized in that, The engine for generating operational data analysis tasks includes: Each analysis task in the operational data analysis task list is sequentially associated with the enterprise operational knowledge graph to obtain a set of knowledge nodes associated with operational analysis tasks. Historical operational data mining is performed based on the set of knowledge nodes associated with the operational analysis tasks to obtain a dataset of knowledge associated with the operational analysis tasks. Predictive training and optimization are performed on the knowledge dataset associated with the operational analysis tasks to construct a set of operational data analysis task models. The operational data analysis task model set is coupled and driven to generate the operational data analysis task engine.

9. The knowledge graph-based enterprise operation data analysis method as described in claim 8, characterized in that, The set of operational data analysis task models includes: Based on the characteristic information of the knowledge dataset associated with the operational analysis task, select the operational analysis task model structure set; The operational data analysis task model set is constructed by using the operational analysis task model structure set to perform label prediction training and cross-validation optimization on the operational analysis task associated knowledge dataset.

10. A knowledge graph-based enterprise operation data analysis system, characterized in that: The system is used to implement the knowledge graph-based enterprise operation data analysis method according to any one of claims 1 to 9, the system comprising: The data acquisition module is used to collect enterprise multi-source heterogeneous operation datasets, perform standardization processing and multi-dimensional entity extraction on the enterprise multi-source heterogeneous operation datasets, and obtain enterprise operation multi-dimensional entity sets. The knowledge graph construction module is used to perform knowledge graph modeling and dynamic update optimization based on the enterprise operation multi-dimensional entity set to construct an enterprise operation knowledge graph. The prediction-driven module is used to construct an operational data analysis task list, perform prediction-driven analysis on the enterprise operational knowledge graph according to the operational data analysis task list, and generate an operational data analysis task engine. The operational decision analysis module is used to acquire data during the enterprise's operation period, perform risk prediction and reasoning and operational decision analysis on the data based on the operational data analysis task engine, determine the target enterprise's operational decision plan, and carry out operational decision control through the target enterprise's operational decision plan.

Images