Optimization of AdaBoost Algorithm by PSO and Its Application in Translation Initiation Sites Prediction

In this paper, we optimize the boosted results of AdaBoost algorithm by particle swarm optimization (PSO) and form a learning algorithm PSO-AB. We use it to boost the classification ability of support vector machine (SVM) and back propagation neural network (BPNN). This PSO-AB adopts SVM and BPNN to classify the experimental data, uses AdaBoost algorithm to boost the classification results, and then uses PSO to optimize the boosted results. Translation initiation sites (TIS) prediction results show that AdaBoost algorithm could improve the accuracy of training set, but for testing set the result is not satisfactory. PSO-AB makes it possible to maximize the testing accuracy of AdaBoost algorithm on the premise of that the accuracy of training set is still exact. PSO-AB will be a more effective method to classification problems compared with AdaBoost algorithm and base learning algorithm.