Exploring Impact of Feature Selection on Classification Models for Detection of Alzheimer’s Disease

Feature collection and Feature removal are two important techniques for dimensionality reduction that reduces the complexity of the model and solves the problem of overfitting. Redundant features and irrelevant features are removed using different methods for selection of features techniques whereas new features are created with feature extraction. In this paper, we compared different feature assortment techniques such as Recursive Feature Elimination (RFE), Boruta algorithm and Random Forest Algorithm on AD-Alzheimer’s- Disease dataset. We have compared the accuracies of different classification algorithm like SVMC, K-NEAREST-NEIGHBOR and Random Forest without selection of features and with feature selection methods like Boruta and RFE. The results shows that random forest applied with Boruta feature selection outperforms SVMC and K-NEAREST-NEIGHBOR with and without feature selection. Moreover, selecting suitable feature selection technique for classification models has a great impact on its performance.