Feature selection and integration algorithm based classification method

The invention relates to a feature selection and integration algorithm based classification method, which is characterized in that the classification method comprises the following steps: (1) using aninformation gain rate and a symmetric uncertainty to calculate a score of each feature of a data set S for the existing data set S, setting a threshold and screening features, deleting features withthe scores less than the threshold, and forming a new data set S'; and (2) using multiple learners to learn the data set S' after feature selection, adjusting parameters of the learners, using the adjusted learners to train unknown data u, calculating the probability set that the unknown data u belongs to each class, using the average method and the weighted voting method to perform integration calculation on the probability set, and obtaining the class that the unknown data u should be classified. The invention reduces errors caused by a single classifier to a certain extent, and improves theaccuracy of the classifica