Graph Matching Algorithm Based On Rough Set

With the rapid development of intelligent information technology,the amount of data is also expanding,and more and more researchers will deal with these massive data as their research object.As a basic structure that expresses the relationship between data,the graph retains the structural information of the data and the dependencies between the various structures.Graph matching,as one of the basic algorithms in graph theory,has important research significance in many fields such as computer vision and pattern recognition.Among them,the accuracy of non-accurate graph matching is lower than that of exact graph matching,and the requirements for matching constraints are also relatively loose.Therefore,non-accurate graph matching such soft computing technology is more widely used in various fields.The typical non-accurate graph matching methods are: map-based editing distance,graph kernel,genetic algorithm and graph-based methods.These methods have the problems of complicated calculation process and low matching accuracy.The paper adopts rough sets for attribute reduction.Without knowing any information in advance,the attribute dimension can be reduced,thus shortening processing time and improving matching efficiency.Based on rough set theory and graph theory,this paper explores and researches rough set graph matching technology by analyzing rough set algorithm for graph data attributes reduction and corresponding improvement strategies.The main research contents are summarized as follows:Firstly,in the feature extraction phase of graph data,the thesis extracts the non-topology features of common nodes and edges and the topological features that can express the relationship between attributes,so as to better describe the structural information between graph data and construct sufficient information.Rough set decision table.Secondly,the paper uses two classifiers of rough sets to construct each classifier,and then uses classifier fusion to construct multiple classifiers.The heuristic reduction algorithm based on the attribute dependency degree is used to reduce the attributes of the base classifier.Then the attribute of the test data is reduced,then the attributes of the test data are extracted and matched with the rule obtained after the reduction.The matching probability eventually completes the matching of multiple graphs.Finally,the attribute-reduction reduction algorithm is optimized,the property is selected by roulette,the appropriate threshold is determined through experiments to determine the window size,and then the window model is used to process the attributes in the decision table again.The incremental attribute reduction algorithm implements graph matching.The experimental results show that the rough set graph matching methods of the two strategies can not only avoid the time-consuming problem of large-dimension data processing,but also guarantee a certain matching accuracy.