Research And Application Of DNA Genetic Algorithm Based On P System In Cluster Analysis

Clustering analysis is significant tool in data mining.A variety of clustering algorithms have appeared in academia and practical work applications,but various clustering algorithms are not perfect.In the research work of this thesis,a novel DNA genetic operator is proposed,and an improved DNA genetic algorithm combined with P system is proposed to optimize fuzzy C-means clustering and density peak clustering.In addition,the improved fuzzy C-means clustering algorithm was applied to actual text clustering.An improved DNA genetic algorithm(IDNA-DMS)based on structural variability of cell-like P system was proposed(Improved DNA-Dynamic Membrane Structure,hereinafter referred to simply as IDNA-DMS).Firstly,a structural variability cell-like P system was designed as an operation framework,which utilized the evolutionary rules of the object and the unique membrane evolution rules of the structural variability cell-like P system: membrane dissolution and membrane creation,and improved computational efficiency.Secondly,based on the biological knowledge of nature,a new operator of DNA genetic algorithm,the splicing operator,is designed.According to the structural characteristics of chromosomes,two different types of splicing operators are designed: inner splicing operator and outer splicing.The new operator enhances the population diversity of the DNA pool during the evolution process,which is beneficial to jumping out of local optimum limitations,which is beneficial to the optimal solution.Finally,eight standard test functions and two Gaussian kernel functions are used to evaluate the performance of the algorithm,which proves the effectiveness of the algorithm.A weighted fuzzy C-means clustering algorithm(WFCM)based on IDNA-DMS is proposed(Weighted FCM,hereinafter referred to simply as WFCM).Firstly,the calculation formula of FCM algorithm is improved,and the problem that FCM is sensitive to isolated and noise points is solved.Combining the newly proposed IDNA-DMS with WFCM,the initial clustering center of the improved fuzzy C-means clustering is obtained by IDNA-DMS undergoing iterative optimization.Finally,the effectiveness of the algorithm is tested on four real datasets,what's more,compared with FCM and WFCM,the superiority of the algorithm can be proved.The FN-DPC based on IDNA-DMS is proposed(Fuzzy Neighborhood-DPC,hereinafter referred to simply as FN-DPC).The IDNA-DMS algorithm is used to optimize the truncation distance dc,and the fuzzy membership degree is used to optimize the local density.The density peak clustering algorithm itself is improved(This algorithm is called FN-DPC + IDNA-DMS algorithm),and the improved algorithm is tested on the UCI data sets and artificial datasets.The experimental results on the UCI datasets are compared with DPC,DBSCAN and K-means algorithms,which proves that the method has better system stability and robustness.The proposed WFCM+IDNA-DMS algorithm is applied to text clustering experiments to further test the application effect of the algorithm in practical work.The text dataset used in the experiment was obtained online from Sogou Lab.The final result proves the feasibility of the WFCM+IDNA-DMS algorithm in practical work.