بهينه سازي الگوريتم خوشه‌بندي C-means با مدل تركيبي PSO-GA

Optimization of C-means clustering algorithm with PSO-GA combined model

Abstract Clustering, one of the important operations at the conclusion of data mining on the data is considered as often clustering as the first step of data mining processes is remembered and one of the widely-used methods in the field, classic C-means clustering algorithm which is a basic approaches to many other clustering methods but problems such as being sensitive to the initial value, to trap in local optimum and convergence time, this algorithm still threatens and one of the solutions that have been considered for this issue, converting the clustering problem into an optimization problem and solve it using optimization algorithms and on the other hand, in recent years a new generation of algorithms have been proposed as metaheuristic and the main purpose of applying this algorithms to increase the speed of convergence in solving large-scale problems and avoid falling into the trap of local optimal and achieve global optimal points so in the process of this research, clustering as a combination of C-means algorithm and each of the five proposed algorithm(PSO,ABC,DE,HS,GA) implemented with three sets of data and with regard to the results obtained, a proposed model (combinatorial model PSO-GA) was presented and in the end, it was concluded that the proposed model in order to improve the C-means algorithm while the criteria within cluster distance could be successful in terms of convergence rate is almost two times better than the initial GA and PSO that this is a great advantage in the clustering, especially the clustering massive data sets.

Abstract Clustering, one of the important operations at the conclusion of data mining on the data is considered as often clustering as the first step of data mining processes is remembered and one of the widely-used methods in the field, classic C-means clustering algorithm which is a basic approaches to many other clustering methods but problems such as being sensitive to the initial value, to trap in local optimum and convergence time, this algorithm still threatens and one of the solutions that have been considered for this issue, converting the clustering problem into an optimization problem and solve it using optimization algorithms and on the other hand, in recent years a new generation of algorithms have been proposed as metaheuristic and the main purpose of applying this algorithms to increase the speed of convergence in solving large-scale problems and avoid falling into the trap of local optimal and achieve global optimal points so in the process of this research, clustering as a combination of C-means algorithm and each of the five proposed algorithm(PSO,ABC,DE,HS,GA) implemented with three sets of data and with regard to the results obtained, a proposed model (combinatorial model PSO-GA) was presented and in the end, it was concluded that the proposed model in order to improve the C-means algorithm while the criteria within cluster distance could be successful in terms of convergence rate is almost two times better than the initial GA and PSO that this is a great advantage in the clustering, especially the clustering massive data sets.