Dimensionality Reduction is the paramount and vital technique for forecast and categorization of high-dimensional data. Feature Selection is a dimensionality diminution method, which is utilized to examine the high dimensional data with the goal to extract useful attributes from them. It serves the dual purpose of decreasing the search space and enhancing the caliber of machine learning technique with respect to time, space and accuracy. Various feature selection techniques are utilized with positive results in vast fields of application. This paper explores the taxonomy of Feature Selection techniques with their pros and cons. Five parameters are deployed for comparison of the existing techniques. These five parameters are: the Relevance and Redundancy of the subset, Accuracy, Time and Memory requirements, The Stability measure, and The Size of selected subset. This study has been conducted in the hope that it would help researchers decide which technique is best suited for their application.
Volume 11 | 10-Special Issue
Pages: 1180-1190
DOI: 10.5373/JARDCS/V11SP10/20192961