MahaEnemy: Enemy-Based Instance Reduction with Mahalanobis Distance

Instance reduction is an essential pre-processing step in instance-based learning. As dataset sizes continue to expand, computational complexity and memory demands increase significantly. Effective data reduction techniques mitigate these challenges, enhancing efficiency without affecting the model's performance. In this paper, an enhancement of our enemy-based instance reduction technique is proposed by applying the Mahalanobis distance metric. The proposed algorithm is evaluated in comparison with the original one using the Euclidean distance. Experimental results demonstrate that even though the Mahalanobis-based algorithm requires slightly more computational time, it achieves higher model accuracy and a better reduction ratio.