Boosting Performance Classification of Multiple Intelligence Learning Styles in Learning Management System Using Feature Extraction Technique

The demand for personalized and adaptive learning according to learner needs and preferences is based on the learner model; learning style is one of the main components that make up the learner model. One learning style based on the cognitive approach is the multiple-intelligence learning style, which has performed well in face-to-face learning. However, the application of multiple intelligence learning styles in e-Iearning still needs to be improved. The main obstacle is the low performance of this learning style detection classification in online learning. This research aims to improve the classification performance of multiple intelligence learning styles detection in online learning by applying feature extraction techniques to the dataset. This research uses data analysis methods, including data collection from learning interactions with the module-based Learning Management System (LMS) owned by the Directorate General of Teachers and Education Personnel, Ministry of Education, Culture, Research and Technology, Republic of Indonesia to form a dataset, using the Multiple Intelligences Inventory to obtain the dominant intelligence class value, data preprocessing, application of SMOTE-Tomek resampling technique to overcome the problem of the imbalanced dataset, applying feature extraction techniques Independent Component Analysis (ICA) and Singular Value Decomposition (SVD), selection of machine learning classifier models namely K-Nearest Neighbor (K-NN), Random Forest (RF), and Support Vector Machine (SVM), model validation using k-fold cross-validation technique, and model evaluation using accuracy, precision, recall, and fl-score metrics. The experimental results show that the SVM classification performance accuracy reaches 83.33 on the dataset using SVD, 80.56 on the dataset using ICA, and RF reaches 80.11 on the dataset using SVD, a promising result for future development. © 2024 IEEE.