Abstract
Biological classification based on gene codon sequence is critical in life science research. This paper aims to improve the classification performance of conventional algorithms by integrating bacterial foraging optimization (BFO) into the classification process. To enhance the searching capability of conventional BFO, we leverage adaptive T-distribution variation to optimize the swimming step size of BFO, which is named TBFO. Different degree of freedom for t-distribution was used according to the iteration process thus to accelerate converging speed of BFO. The parameters of Artificial Neural Network and Random Forest are then optimized through the TBFO thus to enhance the classification accuracy. Comparative experiment is conducted on six standard data set of DNA codon usage frequencies. Results show that, TBFO performs better in terms of accuracy and convergence speed than PSO, WOA, GA, and BFO.