摘要
针对粒子对算法存在过早陷入局部最优导致聚类精度不高以及聚类结果对初始粒子比较敏感等问题,提出了一种新的基于粒子对(PPO)与差分进化(DE)混合算法。该混合算法结合PPO和DE的优点,根据一定的迭代次数在精英粒子对迭代过程中引入DE算法,借助DE算法的全局收敛能力避免PPO算法过早陷入局部最优的缺点,并借助K-means快速聚类的结果和PSO聚类结果初始化粒子位置,提高初始粒子的质量从而提高聚类结果精度。将混合算法应用于真实的基因表达数据,实验结果表明,混合算法比K-means和PPO算法具有更好的聚类结果和稳定性。
In order to solve the problem that particle pair algorithm existed local optimization premature to lower precision and the clustering results were sensitive to initial particle,this paper put forward a new hybrid algorithm based on particle pair optimization（PPO） and differential evolution（DE）.The hybrid algorithm combined the advantages of PPO and DE,and assigned the fast cluster rusult of the K-means and PSO to initialize particle position to improve the quality of the initial particles and improve accuracy of clustering results.It applied the hybrid algorithm to gene expression data.The experiment results indicate that the hybird algorithm obtains better clustering precision and stability than the K-means algorithm and particle pair algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2012年第7期2484-2487,共4页
Application Research of Computers
基金
广西研究生教育创新计划资助项目(2011106020812M57)
关键词
基因聚类
K-MEANS算法
粒子对
差分进化
混合算法
gene clustering
K-means algorithm
particle pair
differential evolution
hybrid algorithm
