<![CDATA[Feature selection is a critical step in enhancing the accuracy and efficiency of machine learning models, particularly in the field of human activity recognition (HAR) using smartphone sensor data. This research explores the use of Binary Particle Swarm Optimization (BPSO) to reduce the dataset's dimensionality while retaining essential information. Out of 90 initial features, BPSO successfully selected 34 optimal features. These features were then used to train various machine learning models, achieving exceptional classification performance. The Random Forest algorithm achieved the highest accuracy at 99,92%, followed by K-Nearest Neighbour (KNN) at 99,08%, Support Vector Machine (SVM) at 98,68%, Multi-Layer Perceptron (MLP) at 96,01%, Decision Tree (DT) at 95,07%, and Naïve Bayes (NB) at 68,69%. The accuracy of all these algorithms exceeded the accuracy of the same algorithms without PSO-based optimization, as reported in previous studies used as baselines, except for Naive Bayes. These findings highlight the effectiveness of BPSO in feature selection for HAR tasks and its capability to improve machine learning model performance in practical applications.]]>
Copyrights © 2025
