The Study Of Graph Matching Based On Probabilistic Neural Network

With the constantly popularization and innovation of the Internet technology,the result has been a rapid increase of the graph data in every field.Efficient query and matching on large-scale graph data is the basis of the large data analysis problems,Graph matching as the basic algorithm is playing an important role.The existing inexact graph matching algorithm has property redundancy and high complexity.An improved graph matching algorithm is proposed in this paper.First,preprocessing graph data,realize to feature extraction and selection,The second,the improved probabilistic neural network is trained using the optimized eigenvectors,completed the design of classifier,Finally,test the accuracy of the classifier with test data.This paper mainly extracts the topological features and non-topological features of the graph,the topological features of the graph can describe the basic structure of the graph,rather than the non-topological features can describe the basic information of the graph,then combine the both,it can draw a distinguishable figure data information.A probabilistic neural network based on competition(CPNN)is proposed.In CPNN,parameters are optimized through genetic algorithms.Therefore,the optimal network parameters are selected to improve the classification performance.In the aspect of optimizing the topology structure of the probabilistic neural network,an improved learning vector quantization(LVQ)algorithm based on supervised learning is proposed,and the network structure is optimized by the LVQ algorithm to improve the classification performance.The experimental results on different data sets show that although the performance of the above two algorithms is good or bad due to different data sets,they can achieve better results.