Mirror image type industry linkage engine system and method based on knowledge graph

Abstract

The invention discloses a mirror image type industry linkage engine system and method based on a knowledge graph, and relates to the technical field of industry linkage engines, and the system comprises a data collection module which collects multi-industry data in multiple modes, and the cleaning accuracy is 90% or above; the knowledge graph construction module is combined from bottom to top and from top to bottom, and is used for distributed storage and block chain evidence storage; the industry mirror image model generation module is used for generating a model through machine learning, and the association accuracy is 95% or above; the linkage analysis and decision module is used for finding a linkage strategy through an intelligent algorithm; and the user interaction module has a graphical interface, is fast in response, can feed back optimization, and supports multi-language and VR / AR display. The system is accurate and efficient in data processing, integrates multi-source data to construct the knowledge graph, is deep and comprehensive in industry analysis, simulates industry relations to assist decision making, is intelligent and reliable in decision making support, provides strategies and evaluation, is convenient and friendly in user experience, improves interactivity through multiple technologies, promotes accurate decision making of enterprises, and promotes collaborative development and innovation of the industry.

Description

Technical Field

[0001] The present invention relates to the technical field of mirror-type industry linkage engines, and in particular to a mirror-type industry linkage engine system and method based on a knowledge graph. Background Art

[0002] In today's era of rapid globalization and informatization, the links between industries are becoming increasingly close and complex. The operation and development of enterprises are no longer limited to their own industries, but are interdependent and mutually influential with many related industries. While this inter-industry linkage brings opportunities, it also brings huge challenges to enterprises and industry managers.

[0003] Traditional industry analysis and decision-making methods are often limited to the internal data and experience of a single industry, lacking a comprehensive and in-depth understanding of cross-industry relationships. For example, in market research, companies often focus on the direct market of their own products or services, and it is difficult to accurately grasp the transmission effect of potential changes in upstream and downstream industries on themselves. When a certain industry experiences technological innovation or market fluctuations, related companies often miss opportunities to respond or respond blindly because they cannot promptly understand the chain reactions that may be triggered in the industrial chain, resulting in waste of resources and reduced competitiveness.

[0004] Existing data processing and analysis technologies are unable to cope with the massive amounts of data from multiple industries. Data comes from a wide range of sources and in various formats, including structured database data, semi-structured web page data, and unstructured text, image, and audio data. Traditional data collection tools are difficult to integrate this data efficiently and accurately, and the data cleaning and conversion process is cumbersome and error-prone, resulting in uneven data quality and unable to provide a reliable basis for in-depth industry analysis.

[0005] Furthermore, the relationships between industries are complex and involve multiple dimensions such as supply chain, diffusion of technological innovation, market competition and cooperation. In the past, there was a lack of effective models and tools to systematically sort out and simulate these relationships, making it difficult to predict the propagation path, impact scope and time lag of industry changes between different industries. For example, when emerging technologies emerge, traditional analysis methods are difficult to accurately predict their penetration speed and degree of change in related industries, as well as their role in reshaping the business model and competitive landscape of traditional industries.

[0006] In addition, in terms of information interaction and decision support, enterprises and industry managers lack intuitive, convenient and intelligent tools. It is impossible to quickly obtain comprehensive industry-related information and accurate analysis results, making it difficult to make scientific decisions in a complex and changing market environment. In summary, there is an urgent need for an innovative mirror-type industry linkage engine system and method based on knowledge graphs to integrate multi-industry data, build accurate industry relationship models, and realize intelligent linkage analysis and decision support, so as to help enterprises and industries achieve sustainable development and enhance competitiveness in a highly correlated economic environment. Summary of the invention

[0007] The present invention proposes a mirror-type industry linkage engine system and method based on knowledge graph to solve the problems mentioned in the above-mentioned prior art.

[0008] In order to achieve the above purpose, the present invention adopts the following technical solution: a mirror-type industry linkage engine system based on knowledge graph, comprising: Data collection module: For data sources in multiple industries, we use web crawler technology, data interface docking and sensor data collection methods to collect structured, semi-structured and unstructured data; through the combination of rule-based cleaning algorithms and machine learning algorithms, we remove duplicate, erroneous and incomplete data to provide a high-quality data foundation for knowledge graph construction; Knowledge graph construction module: It adopts a combination of bottom-up and top-down construction methods. From the bottom up, entities, relationships and attributes are extracted from the collected data through data mining algorithms, including clustering algorithms and association rule mining algorithms. In the entity extraction process, named entity recognition technology is used in combination with the domain ontology knowledge base. Industry mirror model generation module: Based on the constructed knowledge graph, the neural network model and support vector machine model are used to simulate and map the operation modes, business processes, and market interactions of various industries to generate industry mirror models. In the model generation process, the adversarial generative network (GAN) technology is used to optimize the model. Through adversarial training of the generator and the discriminator, the generated industry mirror model reflects the actual situation of the industry. At the same time, the transfer learning method is introduced to transfer the existing experience of similar industry models to the construction of new industry models. Linkage analysis and decision-making module: Receive industry data change information or analysis scenarios set by users, and conduct linkage analysis based on the industry mirror model; use path search algorithms and optimization algorithms to find the optimal linkage path and strategy between industries; User interaction module: interact with users through a graphical interface.

[0009] Furthermore, in the data acquisition module, the machine learning algorithm uses a deep neural network model to learn data features, automatically identify and correct abnormal patterns in the data, including intelligently judging outliers in the data and correcting them through learning a large amount of historical data, and generating a cleaning report during the data cleaning process, recording in detail the amount of data cleaned, the comparison of data before and after cleaning, and the abnormal data situation; When collecting data, different collection strategies are adopted for different data types. For structured data, such as table data in the database, SQL query statements are used for direct extraction; for semi-structured data, including XML and JSON format data, parsing tools are used for parsing and extraction; for unstructured data, such as text, image, and audio data, feature extraction and conversion are performed first, including the use of natural language processing technology to extract keywords and entities from text data, and the use of image recognition algorithms to extract key information from image data, and then collection is performed to ensure that different types of data can be effectively collected and converted into the data format required for knowledge graph construction. In the process of unstructured data collection, a text classification model based on deep learning is used to pre-classify text data, including industry news, technical reports, market reviews and other categories. Then different keyword extraction and entity recognition strategies are adopted according to the category to improve the pertinence and efficiency of data collection. In the data collection process, data encryption technology is adopted, including SSL / TLS encryption protocol, to ensure the security of data transmission and prevent data from being stolen or tampered with.

[0010] Furthermore, in the knowledge graph construction module, top-down integration and optimization are performed based on the ontology framework predefined by industry experts. The constructed knowledge graph contains the concept system of multiple industry fields, with no less than 100,000 entities, no less than 50 types of relationships, rich attribute information, and a knowledge update mechanism. When new data meets the update conditions, including when the data change exceeds the set threshold or new key data is added, the knowledge graph is updated in a timely manner to maintain the timeliness of knowledge. The knowledge graph adopts a distributed storage architecture, including a distributed file system HDFS based on Hadoop, to improve the storage capacity and read-write efficiency of the knowledge graph data. At the same time, blockchain technology is used to store the construction and update process of the knowledge graph to ensure that the data cannot be tampered with and is traceable. Every modification operation of the knowledge graph is recorded in the blockchain ledger, including detailed information such as the modification time, the modifier, and the modification content. During the construction of the knowledge graph module, standardized industry terms and unified naming rules are used to name entities and relationships. By establishing a term mapping table and ontology alignment technology, the problem of inconsistent terminology in different data sources is solved. For example, different names for the same product by different companies are unified into standard names, which improves the accuracy and consistency of the knowledge graph. The term mapping accuracy rate reaches more than 98%. In the ontology alignment process, an algorithm based on semantic similarity is adopted, including calculating the semantic similarity between terms based on the Word2Vec word vector model, combined with manual review, and using knowledge fusion technology to fuse knowledge graphs from different sources, including fusing the public domain knowledge graph with the internal knowledge graph of the enterprise.

[0011] Furthermore, when generating the model, the industry mirror model generation module adopts a multi-dimensional feature representation method, incorporating macroeconomic environment characteristics, policy and regulatory characteristics, and social and cultural characteristics, including GDP growth rate and inflation rate in macroeconomic indicators, changes in industry regulatory policies in policies and regulations, and changes in consumption trends in social culture, so that the generated industry mirror model can more comprehensively reflect the actual situation of the industry, and the comprehensive evaluation indicators of the model, including accuracy, recall rate, and F1 value, are better than traditional single-dimensional models; when incorporating multi-dimensional features, feature engineering technology is used to screen and reduce the dimensions of the features, including the principal component analysis PCA algorithm to remove redundant features, improve the efficiency and accuracy of model training, and use cross-validation methods during the model training process to comprehensively evaluate and optimize the model performance.

[0012] Furthermore, the linkage analysis and decision-making module takes into account the time lag effect between industries when conducting linkage analysis. By establishing a time series model, including the ARIMA model, the propagation law of industry changes in the time dimension is analyzed, and the time interval of linkage reactions between different industries is determined. When formulating linkage strategies, resource input and action sequence are reasonably arranged according to the time lag effect to improve the implementation effect and success rate of the linkage strategy. The error of the time lag effect analysis is controlled within ±5%. In the process of analyzing the time lag effect, the gray system theory is used to pre-process the data, explore the potential laws in the data, improve the prediction accuracy of the time series model, and combine the dynamic programming algorithm to optimize the linkage strategies at different time points to maximize the overall benefit while considering the feedback mechanism between industries.

[0013] Furthermore, a method for implementing a mirror-type industry linkage system based on a knowledge graph comprises the following steps: Data collection steps: determine the multi-industry scope to be monitored, configure data collection parameters according to industry characteristics, start the data collection module, use a variety of collection methods to collect data from various industries, preliminarily organize and clean the collected data, remove errors and duplicate data, store the cleaned data in the data cache area, and wait for the knowledge graph to be built and used. The data collection cycle is flexibly adjusted according to the frequency of industry data updates. During the data collection process, if data acquisition fails or data quality is seriously degraded, the data collection module automatically switches to the backup collection source or adjusts the collection strategy; Knowledge graph construction steps: read data from the data cache, and construct the knowledge graph in a bottom-up and top-down approach. First, extract entities, relationships, and attributes through data mining algorithms to build a preliminary knowledge graph structure. Then optimize and integrate them according to the ontology framework defined by industry experts, standardize the naming and classification of entities and relationships, and continuously verify and adjust during the construction process. After the construction is completed, store the knowledge graph in the knowledge base, establish a knowledge update mechanism, and update the knowledge graph regularly or when there are major data changes. When constructing the knowledge graph, use a distributed computing framework to accelerate the data processing process. The steps for generating an industry mirror model are as follows: read the knowledge graph data from the knowledge base, input it into the machine learning model, and combine it with multi-dimensional feature data, including macroeconomic data and policy and regulatory data, to train and generate an industry mirror model; during the training process, continuously adjust the model parameters to improve the model's accuracy in mapping industry relationships and its ability to predict industry changes. After the model training is completed, perform a performance evaluation, and deploy the model to the application environment after the evaluation indicators meet the requirements for subsequent linkage analysis; during the model generation process, use model fusion technology to fuse multiple different machine learning models, combine the advantages of different models, improve the accuracy and stability of the industry mirror model, and use the early stopping method to prevent overfitting during the model training process.

[0014] Linkage analysis and decision-making steps: receive external trigger events, including industry data updates and user analysis requests, conduct simulation analysis in the industry mirror model based on event information, use intelligent algorithms to find the optimal linkage path and strategy between industries, consider the time lag effect between industries, conduct risk assessment and benefit prediction for different linkage plans, screen out the optimal linkage plan, and output the analysis results in the form of a visual report, including linkage suggestions, risk assessment, benefit prediction, etc., to provide decision-making basis for enterprises or industry managers, and continuously optimize the analysis process and model based on user feedback; when the analysis results show that there are potential risks or crises between industries, the system automatically generates early warning information and sends it to relevant enterprises or industry regulatory departments; User interaction steps: Provide a graphical user interface, where users enter query commands, including querying knowledge graph information of a specific industry and industry linkage analysis results. The interface extracts relevant information from the knowledge base or analysis results according to the commands and displays it to the user in the form of charts and text reports. Users evaluate and provide feedback on the analysis results, and the system optimizes data collection, knowledge graph construction, and analysis algorithms based on the feedback information; the user interaction step uses natural language processing technology to realize voice interaction functions. Users enter query commands through voice, and the system automatically recognizes and processes them to facilitate user operation. It also supports multi-language interaction to meet the needs of users in different regions. The interface uses virtual reality VR or augmented reality AR technology to display knowledge graphs and analysis results, providing users with a more intuitive and immersive interactive experience.

[0015] Furthermore, in the data collection step, if data acquisition fails or data quality seriously deteriorates during the collection process, the data collection module automatically switches to a backup collection source or adjusts the collection strategy; including switching to local backup data or other alternative network data sources when a certain network data source is inaccessible, while reducing the collection frequency or increasing data verification links to ensure the continuity and stability of data collection; when switching the backup collection source, an intelligent switching algorithm is used to evaluate the priority of the backup collection source based on the quality and stability of historical collection data, giving priority to backup collection sources with high quality and good stability, and when adjusting the collection strategy, an adaptive adjustment algorithm is used to automatically adjust the collection parameters according to the cause and degree of data quality degradation.

[0016] Furthermore, in the linkage analysis and decision-making step, when the analysis results show that there are potential risks or crises between industries, the system automatically generates early warning information and sends it to relevant enterprises or industry regulatory departments; the early warning information includes the risk type, the scope of industries that may be affected, and the estimated degree of impact; When generating early warning information, a risk assessment model is used to quantify the risk, including the use of the analytic hierarchy process (AHP) to determine the weights of risk factors, combined with the fuzzy comprehensive evaluation method to conduct a comprehensive evaluation of the risk. When sending early warning information, a multi-channel redundant sending mechanism is adopted, and text messages are sent through multiple different SMS service providers at the same time to avoid the failure of early warning information sending due to a single channel failure.

[0017] Compared with the prior art, the present invention has the following beneficial effects: At the data processing level, the data collection module uses a variety of collection methods and advanced cleaning technologies, with flexible collection frequency and over 90% cleaning accuracy, and can integrate multi-source heterogeneous data to provide high-quality materials for knowledge graph construction. The knowledge graph construction module adopts a hybrid construction method with distributed storage and blockchain evidence storage, with rich entity relationships and timely updates, ensuring the accuracy, completeness and timeliness of the knowledge graph, and effectively solving the problems of data fragmentation and knowledge update lag.

[0018] In terms of industry analysis, the industry mirror model generation module uses machine learning and innovative optimization technologies to achieve an accuracy rate of over 95% in relationship mapping. The multi-dimensional feature representation combined with transfer learning makes the model comprehensive and efficient. It can accurately simulate industry operating mechanisms and relationships, deeply explore potential industry connections and impact paths, and provide a forward-looking perspective for corporate strategic planning and decision-making.

[0019] In the field of decision support, the linkage analysis and decision-making module uses intelligent algorithms and reinforcement learning, takes into account time lag effects and multi-agent interactions, and has a decision recommendation efficiency of over 85%. It can quickly screen the optimal linkage strategy in a complex industry environment, provide detailed risk assessment and benefit forecasts, and help companies accurately allocate resources and control risks, seize market opportunities and improve their ability to respond.

[0020] In terms of user experience, the user interaction module adopts technologies such as natural language processing, multi-language support and virtual reality. The interface responds quickly, voice interaction is convenient, and multi-language services are provided to global users. The VR / AR display is intuitive, which facilitates user query, understanding and feedback. The system is optimized based on feedback, which enhances user participation and satisfaction, promotes continuous improvement and perfection of the system, promotes industry information sharing and collaborative development, and improves the overall industry operation efficiency and innovation capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic block diagram of a mirror-type industry linkage engine system based on knowledge graph proposed by the present invention; Figure 2 This is a schematic block diagram of a mirror-type industry linkage engine method based on knowledge graph proposed in the present invention. DETAILED DESCRIPTION

[0022] 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.

[0023] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0024] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, and it can be the internal connection of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0025] Reference Figure 1-2 : A mirror-type industry linkage engine system based on knowledge graph, including: Data collection module: For data sources in multiple industries, it uses web crawler technology, data interface docking, and sensor data collection methods to collect structured, semi-structured, and unstructured data. The frequency of data collection can be set, ranging from once an hour to once a day. The collected data covers industry business data, market dynamics data, technology development data, etc. It has a data cleaning function. By combining rule-based cleaning algorithms with machine learning algorithms, it can remove duplicate, erroneous, and incomplete data. The cleaning accuracy rate reaches more than 90%, providing a high-quality data foundation for the construction of knowledge graphs. Among them, the machine learning algorithm uses a deep neural network model to learn data features, automatically identify and correct abnormal patterns in the data, such as by learning a large amount of historical data, intelligently judge outliers in the data and make corrections, and a cleaning report can be generated during the data cleaning process, which records in detail the amount of cleaned data, the comparison of data before and after cleaning, and abnormal data.

[0026] Knowledge graph construction module: adopts a combination of bottom-up and top-down construction methods. From the bottom up, entities, relationships and attributes are extracted from the collected data through data mining algorithms, such as clustering algorithms and association rule mining algorithms. In the process of entity extraction, named entity recognition technology is used in combination with the domain ontology knowledge base to improve the accuracy and professionalism of entity recognition. From the top down, integration and optimization are carried out according to the ontology framework predefined by industry experts. The constructed knowledge graph contains the concept system of multiple industry fields, with no less than 100,000 entities, no less than 50 types of relationships, rich attribute information, and a knowledge update mechanism. When new data meets the update conditions, such as the data change exceeds the set threshold or new key data is added, the knowledge graph can be updated in time to maintain the timeliness of knowledge. The knowledge graph adopts a distributed storage architecture, such as a distributed file system (HDFS) based on Hadoop, to improve the storage capacity and read and write efficiency of the knowledge graph data. At the same time, blockchain technology is used to store the construction and update process of the knowledge graph to ensure the data is tamper-proof and traceable. Every modification operation of the knowledge graph is recorded in the blockchain ledger, including detailed information such as modification time, modifier, and modification content.

[0027] Industry mirror model generation module: Based on the constructed knowledge graph, machine learning models such as neural network models and support vector machine models are used to simulate and map the operating modes, business processes, market interactions, etc. of various industries to generate industry mirror models. The model can reflect the correlation between different industries, such as upstream and downstream relationships in the supply chain, technology dependencies, market competition relationships, etc. The accuracy of correlation mapping reaches more than 95%. The model can predict the scope and degree of impact of industry changes on other industries, and the prediction error is within ±10%. In the model generation process, the adversarial generative network (GAN) technology is used to optimize the model. Through adversarial training of the generator and the discriminator, the generated industry mirror model can more realistically and accurately reflect the actual situation of the industry. At the same time, the transfer learning method is introduced to migrate the existing similar industry model experience to the new industry model construction, reduce the model training time and data requirements, and improve the efficiency of model construction.

[0028] Linkage analysis and decision-making module: Receive industry data change information or analysis scenarios set by users, and conduct linkage analysis based on the industry mirror model. Use intelligent algorithms, such as path search algorithms and optimization algorithms, to find the optimal linkage paths and strategies between industries. For example, in the scenario of industrial upgrading, analyze the linkage propagation paths and resource allocation strategies of technological innovation in various industries. The analysis results are output in the form of a visual report, which includes linkage suggestions, risk assessment, benefit forecasts, etc., to provide decision-making basis for enterprises or industry managers, and the efficiency of decision-making suggestions reaches more than 85%. This module uses reinforcement learning algorithms to continuously optimize linkage strategies, adjust algorithm parameters according to environmental feedback (such as actual market changes, feedback on the effects of enterprises implementing linkage strategies), so that the generated linkage strategies can continuously adapt to market dynamics, and use multi-agent systems to simulate the behavior and decision-making process of different industry entities during the analysis process, so as to more realistically reflect the complex interactive relationship between industries and improve the accuracy and practicality of linkage analysis.

[0029] User interaction module: interacts with users through a graphical interface. Users can enter query commands, such as querying related industry information of a specific industry, industry change trend forecasts, etc., and view partial or overall views of the knowledge graph. The interface response time is less than 2 seconds, and feedback and evaluation of linkage analysis results can be provided. The system optimizes analysis algorithms and models based on feedback to improve system performance and user experience. The user interaction module uses natural language processing technology to implement voice interaction functions. Users can enter query commands through voice, and the system automatically recognizes and processes them, which is convenient for user operation. It also supports multi-language interaction to meet the needs of users in different regions. The interface uses virtual reality (VR) or augmented reality (AR) technology to display knowledge graphs and analysis results, providing users with a more intuitive and immersive interactive experience. For example, users can browse the three-dimensional structure of the knowledge graph in a virtual space through VR devices and view the associations between industries.

[0030] In the present invention, the data acquisition module adopts different acquisition strategies for different data types when collecting data. For structured data, such as table data in a database, SQL query statements are used for direct extraction; for semi-structured data, such as XML and JSON format data, parsing tools are used for parsing and extraction; for unstructured data, such as text, image, and audio data, feature extraction and conversion are performed first, such as natural language processing technology is used to extract keywords, entities, etc. for text data, and key information is extracted by image recognition algorithms for image data, and then collected to ensure that different types of data can be effectively collected and converted into the data format required for knowledge graph construction. In the process of unstructured data collection, a text classification model based on deep learning is used to pre-classify text data, such as industry news, technical reports, market reviews, etc., and then different keyword extraction and entity recognition strategies are used according to the category to improve the pertinence and efficiency of data collection, and data encryption technology, such as SSL / TLS encryption protocol, is used in the data collection process to ensure the security of data transmission and prevent data from being stolen or tampered with.

[0031] In the present invention, during the construction process of the knowledge graph construction module, standardized industry terms and unified naming rules are used for naming entities and relationships. By establishing a term mapping table and ontology alignment technology, the problem of inconsistent terms in different data sources is solved, for example, different names for the same product by different companies are unified into standard names, the accuracy and consistency of the knowledge graph are improved, and the accuracy of term mapping reaches more than 98%. In the ontology alignment process, an algorithm based on semantic similarity is used, such as calculating the semantic similarity between terms based on the Word2Vec word vector model, combined with manual review to ensure the accuracy of ontology alignment, and at the same time, knowledge fusion technology is used to fuse knowledge graphs from different sources, such as fusing public domain knowledge graphs with internal enterprise knowledge graphs, enriching the content and depth of the knowledge graph, and improving the comprehensive coverage of the knowledge graph for the industry.

[0032] In the present invention, the industry mirror model generation module adopts a multi-dimensional feature representation method when generating the model. In addition to considering the basic business characteristics of the industry, it also incorporates macroeconomic environment characteristics, policy and regulatory characteristics, social and cultural characteristics, such as GDP growth rate and inflation rate in macroeconomic indicators, changes in industry regulatory policies in policies and regulations, changes in consumption trends in social culture, etc., so that the generated industry mirror model can more comprehensively reflect the actual situation of the industry, and the comprehensive evaluation indicators of the model, such as accuracy, recall rate, F1 value, etc., are better than traditional single-dimensional models. When incorporating multi-dimensional features, feature engineering technology is used to screen and reduce the dimensions of features, such as using the principal component analysis (PCA) algorithm to remove redundant features, improve model training efficiency and accuracy, and use cross-validation methods such as k-fold cross-validation in the model training process to comprehensively evaluate and optimize model performance to ensure the generalization ability and stability of the model.

[0033] In the present invention, the linkage analysis and decision-making module considers the time lag effect between industries when performing linkage analysis. By establishing a time series model, such as an ARIMA model, analyzing the propagation law of industry changes in the time dimension, determining the time interval of linkage reactions between different industries, and formulating linkage strategies, rationally arranging resource inputs and action sequences according to the time lag effect, improving the implementation effect and success rate of the linkage strategy, and the error control of the time lag effect analysis is within ±5%. In the process of analyzing the time lag effect, the gray system theory is used to pre-process the data, the potential laws in the data are excavated, the prediction accuracy of the time series model is improved, and the linkage strategy at different time points is optimized in combination with the dynamic programming algorithm to maximize the overall benefit, while considering the feedback mechanism between industries, such as the secondary impact of the reaction of other industries on the original industry after the changes in one industry have an impact on other industries, so that the linkage analysis is more in line with the actual situation.

[0034] The present invention discloses a mirror-type industry linkage method based on a knowledge graph, comprising the following steps: Data collection steps: determine the multi-industry scope to be monitored, configure data collection parameters according to industry characteristics, start the data collection module, use a variety of collection methods to collect data from various industries, preliminarily organize and clean the collected data, remove obvious errors and duplicate data, store the cleaned data in the data cache, and wait for the knowledge graph to be built and used. The data collection cycle can be flexibly adjusted according to the industry data update frequency, generally between 1 hour and 24 hours. During the data collection process, if data acquisition fails or data quality is seriously degraded, the data collection module automatically switches to the backup collection source or adjusts the collection strategy. For example, when a network data source cannot be accessed, switch to local backup data or other alternative network data sources, and reduce the collection frequency or increase the data verification link to ensure the continuity and stability of data collection. The data collection success rate remains above 90%. In addition, data compression technology, such as the gzip compression algorithm, is used in the data collection process to reduce data storage space and transmission bandwidth requirements and improve data collection efficiency.

[0035] Knowledge graph construction steps: read data from the data cache, and construct the knowledge graph in a bottom-up and top-down approach. First, extract entities, relationships, and attributes through data mining algorithms to build a preliminary knowledge graph structure, and then optimize and integrate them according to the ontology framework defined by industry experts, standardize the naming and classification of entities and relationships, and continuously verify and adjust during the construction process to ensure the integrity and accuracy of the knowledge graph. After the construction is completed, the knowledge graph is stored in the knowledge base, and a knowledge update mechanism is established to update the knowledge graph regularly or when there are major data changes. When constructing the knowledge graph, a distributed computing framework, such as the Spark framework, is used to accelerate the data processing process and improve the speed of knowledge graph construction. In the process of knowledge update, an incremental learning algorithm is used to learn and update only new knowledge, reduce repeated calculations of existing knowledge graphs, and improve knowledge update efficiency.

[0036] The steps to generate the industry mirror model are as follows: read the knowledge graph data from the knowledge base, input it into the machine learning model, and train and generate the industry mirror model by combining multi-dimensional feature data, such as macroeconomic data, policy and regulatory data, etc. During the training process, the model parameters are continuously adjusted to improve the model's mapping accuracy of industry associations and its ability to predict industry changes. After the model training is completed, a performance evaluation is performed. When the evaluation indicators meet the requirements, the model is deployed to the application environment for subsequent linkage analysis. During the model generation process, model fusion technology is used, such as fusing multiple different machine learning models, combining the advantages of different models, and improving the accuracy and stability of the industry mirror model. In addition, the early stopping method is used during the model training process to prevent overfitting. When the performance of the model on the validation set no longer improves, the training is stopped and the optimal model parameters are saved.

[0037] Linkage analysis and decision-making steps: Receive external trigger events, such as industry data updates, user analysis requests, etc., and perform simulation analysis in the industry mirror model based on event information. Use intelligent algorithms to find the optimal linkage path and strategy between industries, consider the time lag effect between industries, conduct risk assessment and benefit prediction for different linkage plans, screen out the optimal linkage plan, and output the analysis results in the form of a visual report, including linkage suggestions, risk assessment, benefit prediction, etc., to provide decision-making basis for enterprises or industry managers, and continuously optimize the analysis process and model based on user feedback. When the analysis results show that there are potential risks or crises between industries, the system automatically generates early warning information and sends it to relevant enterprises or industry regulatory departments. The early warning information includes risk type, the scope of industries that may be affected, and the expected degree of impact. The accuracy of sending early warning information reaches more than 95%, and the sending methods include email, SMS, system message push, etc., so that relevant parties can take timely response measures to reduce risk losses. In the process of linkage analysis, the sensitivity analysis method is used to analyze the model parameters to determine which parameters have a greater impact on the linkage results, so that these parameters can be focused on and adjusted in actual applications to improve the reliability and operability of linkage analysis.

[0038] User interaction step: Provide a graphical user interface, where users can enter query commands, such as querying knowledge graph information of a specific industry, industry linkage analysis results, etc. The interface extracts relevant information from the knowledge base or analysis results according to the command and displays it to the user in the form of charts, text reports, etc. Users can also evaluate and provide feedback on the analysis results. The system optimizes data collection, knowledge graph construction, analysis algorithms, etc. based on the feedback information to improve the system's performance and user satisfaction. The user interaction step uses natural language processing technology to implement voice interaction functions. Users can enter query commands through voice, and the system automatically recognizes and processes them, which is convenient for user operation. It also supports multi-language interaction to meet the needs of users in different regions. The interface uses virtual reality (VR) or augmented reality (AR) technology to display knowledge graphs and analysis results, providing users with a more intuitive and immersive interactive experience. For example, users can browse the three-dimensional structure of the knowledge graph in a virtual space through VR devices and view the correlation between industries. In the process of user feedback processing, machine learning algorithms are used to mine and analyze user feedback data, automatically extract key information on user needs and opinions, and provide a more targeted basis for system optimization.

[0039] In the present invention, in the data collection step, if data acquisition fails or data quality seriously deteriorates during the collection process, the data collection module automatically switches to an alternate collection source or adjusts the collection strategy. For example, when a certain network data source cannot be accessed, it switches to local backup data or other alternative network data sources, while reducing the collection frequency or increasing the data verification link to ensure the continuity and stability of data collection, and the data collection success rate remains above 90%. When switching the alternate collection source, an intelligent switching algorithm is used to evaluate the priority of the alternate collection source based on the quality and stability of historical collection data, and priority is given to alternate collection sources with high quality and good stability. When adjusting the collection strategy, an adaptive adjustment algorithm is used to automatically adjust the collection parameters according to the cause and degree of data quality degradation, such as adjusting the crawling depth of the web crawler, the request frequency of the data interface, etc., to improve the adaptability and flexibility of data collection.

[0040] In the present invention, in the linkage analysis and decision-making step, when the analysis results show that there are potential risks or crises between industries, the system automatically generates early warning information and sends it to relevant enterprises or industry regulatory departments. The early warning information includes the risk type, the scope of industries that may be affected, the expected degree of impact, etc. The accuracy of sending the early warning information reaches more than 95%, and the sending methods include e-mail, text messages, system message push, etc., so that the relevant parties can take countermeasures in time to reduce risk losses. When generating early warning information, a risk assessment model is used to quantitatively evaluate the risk, such as using the analytic hierarchy process (AHP) to determine the weight of risk factors, and combining the fuzzy comprehensive evaluation method to comprehensively evaluate the risk, improve the accuracy and scientificity of the early warning information, and when sending the early warning information, a multi-channel redundant sending mechanism is used to ensure that the early warning information can be delivered to the relevant parties in a timely and accurate manner, such as sending text messages through multiple different SMS service providers at the same time to avoid the failure of early warning information sending due to a single channel failure.

[0041] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A mirror-type industry linkage engine system based on knowledge graph, characterized in that: include: Data collection module: For data sources in multiple industries, we use web crawler technology, data interface docking and sensor data collection methods to collect structured, semi-structured and unstructured data; through the combination of rule-based cleaning algorithms and machine learning algorithms, we remove duplicate, erroneous and incomplete data to provide a high-quality data foundation for knowledge graph construction; Knowledge graph construction module: It adopts a combination of bottom-up and top-down construction methods. From the bottom up, entities, relationships and attributes are extracted from the collected data through data mining algorithms, including clustering algorithms and association rule mining algorithms. In the entity extraction process, named entity recognition technology is used in combination with the domain ontology knowledge base. Industry mirror model generation module: Based on the constructed knowledge graph, the neural network model and support vector machine model are used to simulate and map the operation modes, business processes, and market interactions of various industries to generate industry mirror models. In the model generation process, the adversarial generative network (GAN) technology is used to optimize the model. Through adversarial training of the generator and the discriminator, the generated industry mirror model reflects the actual situation of the industry. At the same time, the transfer learning method is introduced to transfer the existing experience of similar industry models to the construction of new industry models. Linkage analysis and decision-making module: Receive industry data change information or analysis scenarios set by users, and conduct linkage analysis based on the industry mirror model; use path search algorithms and optimization algorithms to find the optimal linkage path and strategy between industries; User interaction module: interact with users through a graphical interface.

2. The mirror-type industry linkage engine system based on knowledge graph according to claim 1 is characterized in that: In the data acquisition module, the machine learning algorithm uses a deep neural network model to learn data features, automatically identify and correct abnormal patterns in the data, including intelligently judging outliers in the data and correcting them through learning a large amount of historical data. In addition, a cleaning report can be generated during the data cleaning process, recording in detail the amount of data cleaned, the comparison of data before and after cleaning, and the abnormal data situation; When collecting data, different collection strategies are adopted for different data types. For structured data, such as table data in the database, SQL query statements are used for direct extraction; for semi-structured data, including XML and JSON format data, parsing tools are used for parsing and extraction; for unstructured data, including text, image, and audio data, feature extraction and conversion are performed first, including the use of natural language processing technology to extract keywords and entities from text data, and the use of image recognition algorithms to extract key information from image data, and then collection is performed to ensure that different types of data can be effectively collected and converted into the data format required for knowledge graph construction. In the process of unstructured data collection, a text classification model based on deep learning is used to pre-classify text data, including industry news, technical reports, market reviews and other categories, and then different keyword extraction and entity recognition strategies are adopted according to the category to improve the pertinence and efficiency of data collection. In the data collection process, data encryption technology is used, including SSL / TLS encryption protocol, to ensure the security of data transmission and prevent data from being stolen or tampered with.

3. The mirror-type industry linkage engine system based on knowledge graph according to claim 1 is characterized in that: In the knowledge graph construction module, top-down integration and optimization are carried out based on the ontology framework predefined by industry experts. The constructed knowledge graph contains the concept system of multiple industry fields, with no less than 100,000 entities, no less than 50 types of relationships, rich attribute information, and a knowledge update mechanism. When new data meets the update conditions, including when the data change exceeds the set threshold or new key data is added, the knowledge graph is updated in time to maintain the timeliness of knowledge. The knowledge graph adopts a distributed storage architecture, including a distributed file system HDFS based on Hadoop, to improve the storage capacity and read-write efficiency of the knowledge graph data. At the same time, blockchain technology is used to store the construction and update process of the knowledge graph to ensure that the data cannot be tampered with and is traceable. Every modification operation of the knowledge graph is recorded in the blockchain ledger, including detailed information such as modification time, modifier, and modification content. During the construction of the knowledge graph module, standardized industry terms and unified naming rules are used to name entities and relationships. By establishing a term mapping table and ontology alignment technology, the problem of inconsistent terminology in different data sources is solved. For example, different names for the same product by different companies are unified into standard names, which improves the accuracy and consistency of the knowledge graph. The term mapping accuracy rate reaches more than 98%. In the ontology alignment process, an algorithm based on semantic similarity is adopted, including calculating the semantic similarity between terms based on the Word2Vec word vector model, combined with manual review, and using knowledge fusion technology to fuse knowledge graphs from different sources, including fusing the public domain knowledge graph with the internal knowledge graph of the enterprise.

4. The mirror-type industry linkage engine system based on knowledge graph according to claim 1 is characterized in that: When generating the model, the industry mirror model generation module adopts a multi-dimensional feature representation method, incorporates macroeconomic environment characteristics, policy and regulatory characteristics, and social and cultural characteristics, including GDP growth rate and inflation rate in macroeconomic indicators, changes in industry regulatory policies in policies and regulations, and changes in consumption trends in social culture, so that the generated industry mirror model can more comprehensively reflect the actual situation of the industry. The comprehensive evaluation indicators of the model, including accuracy, recall rate, and F1 value, are better than traditional single-dimensional models. When incorporating multi-dimensional features, feature engineering technology is used to screen and reduce the dimensions of the features, including the principal component analysis PCA algorithm to remove redundant features, improve the efficiency and accuracy of model training, and use a cross-validation method during the model training process to comprehensively evaluate and optimize the model performance.

5. The mirror-type industry linkage engine system based on knowledge graph according to claim 1 is characterized in that: The linkage analysis and decision-making module takes into account the time lag effect between industries when conducting linkage analysis. By establishing a time series model, including the ARIMA model, the propagation law of industry changes in the time dimension is analyzed to determine the time interval of linkage reactions between different industries. When formulating linkage strategies, resource input and action sequence are reasonably arranged according to the time lag effect to improve the implementation effect and success rate of the linkage strategy. The error of the time lag effect analysis is controlled within ±5%. In the process of analyzing the time lag effect, the gray system theory is used to pre-process the data, explore the potential laws in the data, improve the prediction accuracy of the time series model, and combine the dynamic programming algorithm to optimize the linkage strategies at different time points to maximize the overall benefit while considering the feedback mechanism between industries.

6. A method for realizing a mirror-type industry linkage system based on a knowledge graph, characterized in that: The following steps are involved: Data collection steps: determine the multi-industry scope to be monitored, configure data collection parameters according to industry characteristics, start the data collection module, use a variety of collection methods to collect data from various industries, preliminarily organize and clean the collected data, remove errors and duplicate data, store the cleaned data in the data cache area, and wait for the knowledge graph to be built and used. The data collection cycle is flexibly adjusted according to the frequency of industry data updates. During the data collection process, if data acquisition fails or data quality is seriously degraded, the data collection module automatically switches to the backup collection source or adjusts the collection strategy; Knowledge graph construction steps: read data from the data cache, and construct the knowledge graph in a bottom-up and top-down approach. First, extract entities, relationships, and attributes through data mining algorithms to build a preliminary knowledge graph structure. Then optimize and integrate them according to the ontology framework defined by industry experts, standardize the naming and classification of entities and relationships, and continuously verify and adjust during the construction process. After the construction is completed, store the knowledge graph in the knowledge base, establish a knowledge update mechanism, and update the knowledge graph regularly or when there are major data changes. When constructing the knowledge graph, use a distributed computing framework to accelerate the data processing process. The steps of generating the industry mirror model are as follows: the knowledge graph data is read from the knowledge base, input into the machine learning model, and combined with multi-dimensional feature data, including macroeconomic data and policy and regulatory data, to train and generate the industry mirror model; during the training process, the model parameters are continuously adjusted to improve the model's mapping accuracy of industry associations and its ability to predict industry changes. After the model training is completed, performance evaluation is performed. After the evaluation indicators meet the requirements, the model is deployed to the application environment for subsequent linkage analysis; during the model generation process, model fusion technology is used to fuse multiple different machine learning models, combining the advantages of different models to improve the accuracy and stability of the industry mirror model, and early stopping is used to prevent overfitting during the model training process; Linkage analysis and decision-making steps: receive external trigger events, including industry data updates and user analysis requests, conduct simulation analysis in the industry mirror model based on event information, use intelligent algorithms to find the optimal linkage path and strategy between industries, consider the time lag effect between industries, conduct risk assessment and benefit prediction for different linkage plans, screen out the optimal linkage plan, and output the analysis results in the form of a visual report, including linkage suggestions, risk assessment, benefit prediction, etc., to provide decision-making basis for enterprises or industry managers, and continuously optimize the analysis process and model based on user feedback; when the analysis results show that there are potential risks or crises between industries, the system automatically generates early warning information and sends it to relevant enterprises or industry regulatory departments; User interaction steps: Provide a graphical user interface, where users enter query commands, including querying knowledge graph information of a specific industry and industry linkage analysis results. The interface extracts relevant information from the knowledge base or analysis results according to the command and displays it to the user in the form of charts and text reports. Users evaluate and provide feedback on the analysis results, and the system optimizes data collection, knowledge graph construction, analysis algorithms, etc. based on the feedback information; The user interaction step uses natural language processing technology to realize the voice interaction function. Users input query commands through voice, and the system automatically recognizes and processes them to facilitate user operation. It also supports multi-language interaction to meet the needs of users in different regions. The interface uses virtual reality VR or augmented reality AR technology to display knowledge graphs and analysis results, providing users with a more intuitive and immersive interactive experience.

7. According to the method of the mirror-type industry linkage system based on the knowledge graph in claim 6, in the data collection step, if the data acquisition fails or the data quality is seriously degraded during the collection process, the data collection module automatically switches to a backup collection source or adjusts the collection strategy; This includes switching to local backup data or other alternative network data sources when a certain network data source is inaccessible, while reducing the collection frequency or increasing data verification steps to ensure the continuity and stability of data collection; When switching backup collection sources, an intelligent switching algorithm is used to evaluate the priority of backup collection sources based on the quality and stability of historical collection data, and priority is given to backup collection sources with high quality and good stability. When adjusting the collection strategy, an adaptive adjustment algorithm is used to automatically adjust the collection parameters according to the cause and degree of data quality degradation.

8. According to the method of the mirror-type industry linkage system based on the knowledge graph in claim 6, in the linkage analysis and decision-making step, when the analysis results show that there are potential risks or crises between industries, the system automatically generates early warning information and sends it to relevant enterprises or industry regulatory departments; the early warning information includes the risk type, the scope of industries that may be affected, and the expected degree of impact; When generating early warning information, a risk assessment model is used to quantify the risk, including the use of the analytic hierarchy process (AHP) to determine the weights of risk factors, combined with the fuzzy comprehensive evaluation method to conduct a comprehensive evaluation of the risk. When sending early warning information, a multi-channel redundant sending mechanism is adopted, and text messages are sent through multiple different SMS service providers at the same time to avoid the failure of early warning information sending due to a single channel failure.

Images