<![CDATA[Conventional student performance evaluation mechanisms often fail to detect a decline in performance in a timely manner because they overlook non-academic factors. This study aims to build a comprehensive student performance classification model by simultaneously integrating academic performance and lifestyle patterns using the k-Nearest Neighbor (k-NN) algorithm. The research method utilizes 1,000 student datasets from Kaggle, which were divided using the Stratified Split method into 80% training data and 20% test data. Simulations were run using RapidMiner Studio software with the Euclidean Distance metric and varying k values of 3, 5, and 7. Experimental results show the model achieved a peak global accuracy of 97% at the optimal k value of 5, where the model successfully classified 194 out of 200 test data samples correctly. Lifestyle attributes independently influence students’ academic performance, with the number of study hours per day being the most significant predictor (47.94%). Sensitivity analysis confirms that k=5 is the optimal value to avoid data noise at k=3 (94.50%) and oversmoothing at k=7. In conclusion, this k-NN-based model is highly suitable for application as an early warning system for schools to provide proactive and personalized pedagogical interventions.]]>
Copyrights © 2026
