A Comparative Study of Feature Selection Methods for Activity Recognition in the Smart Home Environment

Feature selection plays an essential role in machine learning for reducing irrelevant, noisy, and redundant features, and selecting the optimal set of attributes for creating compelling predictive models of the study dataset subject. The feature selection (FS) methods entail deciding which essential features to use in machine learning (ML) for model development, and removing redundant features which decrease the overall classification accuracy. This study focused on a comparative analysis of three popular FS approaches: filter, wrapper, and embedded. In this regard, we selected a single technique from each approach, filter-based (Information Gain (IG)), wrapper–based (Recursive Feature Elimination (RFE)), and embedded–based (Tree-based) feature selections. Thereafter, the study applied five base learner classifiers: Logistic Regression (LR), Linear Discriminant Analysis (LDA), Naïve Bayes (NB), Random Forest (RF), and K-nearest neighbor (KNN), to build the human activity recognition (HAR) model. The experimental results show that RF outscored compared to the other classifiers when applied under tree-base feature selection, with an accuracy of 98.9% in house A, and 99.8% in house B. The FS methods enhanced forecast accuracy in the ARAS dataset more than before FS.