基于改进遗传算法的模糊聚类研究及应用

Research and Application of Fuzzy Clustering Based on Improved Genetic Algorithm

模糊C-均值聚类算法是一种局部搜索算法,采用迭代的爬山技术,对初值敏感易陷入局部最小值。遗传算法是一种全局优化算法,能够克服模糊C-均值聚类算法陷入局部最小值的问题,但遗传算法收敛速度慢,易早熟。应用小生境思想对遗传算法进行了改进,以保护种群中基因的多样性,设计了基于最短距离的算术交叉算子、边界变异算子及双精英种子参与进化的策略。仿真实验结果表明,改进后的算法能够提高模糊聚类的收敛速度和聚类质量。

Fuzzy C-means clustering algorithm is an iterative hill-climbing technique for the local search algo- rithm, due to the sensitive dependence on initial conditions and easy to fall into the local minimum. Genetic algo- rithm is a global optimization algorithm, can overcome the fuzzy C- means clustering algorithm to fall into the local minimum problem, but the genetic algorithm has slow convergence, premature convergence. Application of niche theory on genetic algorithm improvements, design based on shortest distance arithmetic crossover operator, mutation operator, boundary double elite seed in evolutionary strategy, in order to protect the population genetic diversity. The simulation results show that, the improved algorithm can improve the convergence speed of fuzzy clustering and clustering quality.