基于改进的和声搜索算法的特征基因选择

Feature Gene Selection Based on Improved Harmony Search Algorithm

针对基因表达谱高维、小样本、高噪声及高冗余等特点,提出一种基于改进的和声搜索算法的特征基因选择方法。首先,采用Kruskal-Wallis算法对原始基因进行初选,降低和声算法搜索空间维数,保证和声搜索算法的优化精度和收敛速度;然后,针对和声搜索算法易陷入局部最优问题,对当前种群中最优、最差和声分别进行进化;同时融合教与学优化算法中个体更新方式,设计一种改进的和声搜索算法实现特征基因选择。仿真实验结果表明,方法在优化精度、时间效率和稳定性等方面优于HS、IHS、EHS和GHS等算法。

A feature gene selection method was proposed based on improved harmony search algorithm aiming at the characters of high-dimension,small samples,high noise and high redundancy of gene expression profile. Firstly,Kruskal-Wallis algorithm is used to select the some genes in order to reduce the dimension of the search space and guarantee the optimization precision and convergence speed of harmony search algorithm. Then,the optimal and the worst harmonics are evolved respectively and the updating approach of individual in teaching-learning-based optimization is integrated to harmony search algorithm at the same time. Simulation results show that the proposed method outperforms HS and improved algorithms,such as IHS,EHS and GHS in terms of optimization accuracy,time efficiency and stability.