Improved Research And Application Of Fuzzy C Mean Algorithm

With the era of big data coming,a technique for data processing through computer algorithms-data mining emerged.Clustering analysis is an unsupervised method in data mining,which can automatically cluster from sample data,and no classification criteria need to be determined before analysis.Because of the above characteristics of clustering analysis,it has been widely applied in many fields.These characteristics of clustering analysis have made them widely used in many fields.In this paper,the improved fuzzy C mean algorithm IFCM(Improved Fuzzy C-means)is improved for the traditional fuzzy C mean algorithm,which is sensitive to the initial center,the clustering number is difficult to determine,the speed of convergence is slow and the local optimality is easy to be trapped.First of all,the paper starts from the angle of information granularity.Based on the principle of particle size analysis,the clustering validity function is constructed according to the cohesion and coupling degree of information granularity,and the clustering results are evaluated through the validity function.Then,the initial clustering center is selected by point density function,which improves the deficiency of traditional algorithm to initialize cluster centers randomly,which makes the convergence speed of the algorithm faster and improves the effectiveness of the clustering results.Aiming at the instability of cluster number,the method of merging cluster centers by successive decrement of category number is adopted.Finally,the best clustering results are obtained according to the clustering validity function.Through the simulation experiment,it is proved that the improved algorithm makes up for the shortcomings of traditional FCM algorithm,such as sensitivity to initial value,instability of clustering class and slow convergence rate.The classification results of IFCM algorithm are more reasonable,which improves the effectiveness of clustering algorithm.The paper further applies the improved algorithm to the classification of load characteristics of substation,and has obtained a better classification effect,which provides a useful reference for the decision of power load application.