Identification and Prediction of Multiple Intelligence Patterns Among University Students Using Machine Learning for Personalized Learning and Instruction

Abstract

Humans possess seven types of intelligence for learning. However, conventional educational systems mostly rely only on the Linguistic and Mathematical types. This narrow focus leaves other intelligences unrecognized and unutilized in learners’ learning processes. The main challenge lies in holistically addressing diverse intelligence levels in typical educational settings. Prior research isolated dominant intelligence traits for personalization, which lack objectivity and fail to utilize the full spectrum of learning potentials. The goal of this research is to identify and predict underlying learning patterns in university students, enabling addressing all types of intelligence holistically at the group level. To achieve this, a multi-algorithm framework integrating unsupervised and supervised learning is developed to analyze Multiple Intelligences profiles and associated learner parameters from the Learning Meta-Learning dataset. MI profiles capture all seven types of intelligences. K-Means clustering identified three unique and interpretable MI profile patterns, outperforming the Two-Step clustering algorithm. For prediction, CatBoost achieved 92.20% accuracy across seven algorithms, including Random Forest, Extra Trees, XGBoost, LightGBM, K-Nearest Neighbors, and HistGradientBoosting. By incorporating Explainable Artificial Intelligence through Shapley Additive Explanations and optimizing features, its accuracy improved to 93.66%. The identified Multiple Intelligences profile patterns offer actionable insights. Beyond the high-level guidelines for personalizing instruction presented in this paper, they can inform various approaches to personalizing education, including adaptive learning module design, curriculum planning, academic counseling, and resource allocation, by providing a structured understanding of learner diversity and a predictive framework that facilitates future research.