多目标人工蜂群双聚类算法在基因表达数据中的应用研究

Research and Application of Multi-Objective Artificial Bee Colony Biclustering in Gene Expression Data

基于多目标优化的双聚类算法能够同时优化均方残差和尺寸等多个相互冲突的目标,更好地挖掘出均方残差较小、尺寸较大的双聚类,提出了一个多目标人工蜂群双聚类算法.该方法首先采用组信息对蜜源进行编码,然后使用2种交叉和1种变异操作分别实现算法的局部搜索和全局搜索,最后根据非劣排序和拥挤距离对外部档案进行修剪.在2套真实的基因表达数据集上进行实验,结果表明:与其他公开算法相比,多目标人工蜂群双聚类算法具有较好的收敛性和种群多样性,同时挖掘出具有显著生物意义的双聚类.

Biclustering algorithms based on multi-objective optimization,which can optimize several objectives simultaneously in conflict with each other,such as the mean squared residue and the size. In order to mine better biclusters with lower mean squared residue but larger size,a novel algorithm named Multi-objective Artificial Bee Colony Biclustering is proposed. Firstly,the approach adopts a group based representation for the genes-conditions associations to encode foods,then two different crossovers and a mutation operation are used to realize local search and global search respectively. Consequently,the non-dominated sort and crowding distance are applied to prune external archives. Experiments are performed on two real gene expression datasets,and it is found that compared with competing algorithms,the method has better global astringency and diversity of the population. Besides,it can obtain significantly biological biclusters.