A hybrid evolutionary algorithm for promoter recognition

The aim of this study is to identify the smallest possible set of features and optimal model parameters for promoter recognition. Particularly, we propose a novel hybrid evolutionary algorithm, which integrates Markov blanket-embedded genetic algorithm (MBEGA), comprehensive learning particle swarm optimize (CLPSO), and support vector machine (SVM) as a whole. This method adopts MBEGA for promoter feature selection while employs CLPSO to optimize the parameters of the promoter identification model. Empirical results on the eukaryotic promoter database (EPD) suggest that, our proposed approach is able to obtain better or competitive classification accuracy than other methods and it is effective and efficient in eliminating irrelevant and redundant features in training process.