An Efficient Feature Selection Approach using Sensitivity Analysis for Machine Learning based Heart Desease Classification

Feature Selection in machine learning is an important pre-processing task. The selection process involves selecting the attribute subset in the original selection set. It tries to identify and eliminate as much information as possible that is irrelevant and redundant. Redundant features help to incorrectly classify data. The removal of redundant characteristics thus reduces data size and computational complexity. It is non-trivial task to identify a good subset of features for effective classification. This paper focuses on the use of an analysis of feature sensitivity to determine the optimum feature subset using MATLAB with improved accuracy and sensitivity for classification. In comparison to the already well-known algorithms for wrapper selection, filter and embedded method, the effectiveness of the proposed algorithms is evaluated. The proposed approach to the selection of features using the Variance based Sensitivity (VSA) approach outperforms the wrapper selection with accuracy 87% and sensitivity 90.12%.