Combination of PSO and AdaBoost algorithm and its application to Multi-Classification prediction of Yeast Protein Localization Sites

A new method PAdaBoost has been proposed to optimize the training sample of AdaBoost algorithm by using PSO, and increased the initial weights of rare class. PAdaBoost model has been used to classify the data of Yeast Protein Localization Sites, for all the data, it has improved the prediction accuracy to 66.24%, up from 59.1% and 64.89% when the best model of and AdaBoost are applied. For studying the generalization of the model, the data set is divided into a training set and a test set, and compared with the AdaBoost model, the accuracy of all data and F-value of rare class are both higher than that of PAdaBoost model. By the model validation, PAdaBoost has shown the stability generalization.