Exploration and Exploitation of High Dimensional Biological Datasets Using a Wrapper Approach Based on Strawberry Plant Algorithm

This paper presents a wrapper approach based on Strawberry Plant Algorithm (SPA) for gene selection in high dimension data classification problem by selecting the most relevant genes for each biological dataset. In order to perform an integrated exploration-exploitation approach to deal the near-optimal (small) gene subset problem obtained from high dimensional microarray data. First, a statistical filter is proposed for gene selection. After, a SPA is proposed to find the most informative genes from the previous gene selection, SPA is applied to explore and exploit new regions of this search and overall to overcome premature convergence. Empirical studies based in five public DNA-microarray datasets it is observed that our model gets the best performances using a smaller number of selected genes than other methods reported in the literature recently.