Preordonance correlation filter for feature selection in the high dimensional classification problem
Hasna Chamlal, Tayeb Ouaderhman, Fadwa Aaboub · 2021
Feature selection is a crucial pre-processing phase in the analysis of high-dimensional datasets. Due to the presence of insignificant and redundant features in the dataset, the classification performance of learning algorithms is generally not satisfactory. The main purpose of feature selection is to surmount the high dimensionality problem by saving relevant features and removing irrelevant and redundant ones from the original features. This research proposes a novel filter feature selection algorithm, termed Preordonance Correlation Filter (PCF), utilized to pinpoint the most discriminating features from the high-dimensional dataset. The performance of the proposed PCF approach is investigated on three artificial datasets of high dimensions. Experimental results demonstrate that the PCF algorithm is able to successfully identify the right significant features and discard the irrelevant ones from all of the three simulated datasets.