Analysis of feature selection technique for human activity recognition

Abstract

The process of Human Activity Recognition using mobile phones is quite complicated, with many extracted features, some of which are redundant. Removing redundant features not only reduces the size of the dataset but also saves time. As a result, our key study aimed to identify the most effective and important features. We propose a noble feature selection technique using Filter method and Wrapper method. On the UCI-HAR dataset, we present a new feature selection approach. For classification, we use the Support Vector Machine, Logistic Regression, Random Forest, K-Nearest Neighbour and Decision Tree classifier in original dataset. The goal is to assess each classifier's performance with a reduced feature set and examine the impact of feature selection on model performance. It is observed that SVM and Random Forest show significant gain in accuracy with the reduced feature set.