Mohammad Taher Horzadeh - Master's degree in software engineering
Ali Akbar Niknafs - Faculty member of Shahid Bahonar University

Growing needs for scalable and efficient feature selection methods prove that existing methods are likely inadequate. This article provides a three-phase approach for feature selection. First, two filter methods including Joint Mutual Information (JMI) and Fisher-score are used. This phase helps improving the classification performance by removing redundant and unimportant features. Second, by combining the results of the previous phase, the obtained features will be intersected. A wrapper method has been used in the third phase with the sequential forward selection and sequential backward elimination. This phase helps selecting relevant feature subset that produce maximum accuracy according to the underlying classifier. Finally, the k nearest neighborhood used to evaluate the classification accuracy of our approach. The empirical results of commonly-used datasets from the UCI repository showed that the proposed method performs better in terms of classification accuracy, number of selected features, and computational complexity.

feature selection, machine learning, data mining, classification, filter, wrapper.