Research On Patent Semantic Community Detection Method Based On Game Theory

With the rapid development of knowledge economy and technology economy,the data in the core scientific and technological resources with patent documents as the main body are characterized by rich knowledge,mass,sparsity,and semantic correlation,which directly promotes the technological development of a country or an enterprise,and provides a reliable basis for guiding technological evaluation.Such technical,massive and rich patent literature data has high technical and economic value.However,due to the lack of patent knowledge deep mining methods,it is difficult for decision-makers to capture knowledge structure information.The demand of decisionmakers for patent analysis,knowledge correlation mining and knowledge structure detection has been more and more urgent.To make patent literature data effectively guide decision makers to conduct technology evaluation and prediction,the research faces the following challenges: 1)Patent literature data is massive and sparse,so it is difficult to mine knowledge information and internal correlations from the global perspective;2)The technical knowledge is rich and mixed,so it is difficult to distinguish the different levels of knowledge;3)The knowledge nodes in the patent network interact with each other,so it is difficult to obtain a stable knowledge structure.At present,the researches on patent technology are mostly based on external features,macro network and knowledge extraction,among which,the existing community detection methods are mostly based on static networks and classical algorithms.They not only lack the hierarchical expression of patent knowledge semantics,but also fails to consider the interaction of nodes during community formation.In order to establish the hierarchical system of patent knowledge and detect its community structure,this thesis proposes a patent semantic community detection method based on game theory and specifically carries out the following research directions.1)In order to mine the technical knowledge and its internal correlation in patent literature and identify the different levels of knowledge in the patent network,the construction method of multi-layer patent knowledge association network is proposed.Firstly,the method builds a patent knowledge network through keyword extraction and association rule mining,mines the inherent semantic associations and enhances the knowledge semantic expression ability of the text.Secondly,the method divides the knowledge keywords into hierarchies,and establishes the pyramid knowledge association structure through entropy calculation.2)In order to implement semantic community detection,an overlapping community detection algorithm based on game theory is proposed.Based on the framework of game theory,the algorithm studies the individual strategies and interactions among intelligent nodes,forms relatively stable semantic communities,and detects the overlapping community structure in the patent knowledge network under the condition of the local equilibrium of the individuals.This thesis combines game theory,community detection and patent knowledge analysis method,and takes individual nodes of patent network as the research perspective to complete community detection in hierarchical knowledge network.The research results are helpful to the analysis of core technology positioning,technology trend prediction and technology evolution structure.