Full Model Selection (FMS) selects the optimal amalgamation of pre-processing technique, feature subset and learning algorithm that obtains the least classification error for a given dataset. Meta-heuristic optimization algorithms are quite suitable for FMS, since it needs to explore and exploit a large solution space. This paper investigates the ability of an efficient meta-heuristic, named Bat algorithm for FMS. Traditional Bat algorithm has been modified and applied for FMS in gene expression analysis. Experiments are conducted on Gene Expression benchmark datasets that shows the suitability and effectiveness of the proposed approach in FMS.