FEATURE SELECTION FRAMEWORK BASED ON FILTER MEASURES FOR HIGH DIMENSIONAL DATA

The increase of data volume in terms of number of features and instances becomes an immense challenge for feature selection algorithms. It increases the computational cost and decreases the accuracy of learning algorithms. This paper proposes a Feature Selection Comprehensive Framework (FSCF) based on filter measures for high dimensional data to produce optimal feature subset in efficient time. Extensive experiments are carried out to comparison of proposed framework and representative methods with respect to different classifiers like Naive bayes and K-NN classifiers on high dimensional datasets. The results demonstrate that proposed framework not only efficient in computational time but also improve the performance of learning algorithms.