Author:
Hameed Sipan M.,Ahmed Walat A.,Othman Masood A.
Abstract
Early and accurate diagnosis of leukemia is crucial for effective treatment. Machine Learning (ML) offers promising tools for leukemia diagnosis classification, but the required high-dimensional datasets pose challenges. This study explores the effectiveness of ML algorithms for leukemia disease classification and investigates the impact of feature selection with the Minimum Redundancy Maximum Relevance (MRMR ) technique. MRMR was implemented to select informative features and evaluate four ML algorithms (Naïve Bayes (NB), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Artificial Neural Networks (ANNs)) using feature subsets with varying levels of relevance based on MRMR scores. Our results demonstrate that MRMR effectively reduced dimensionality while maintaining and even improving classification accuracy. KNN and SVM achieved the highest accuracy (100% for 67, 30, and 24 feature subsets), suggesting the benefit of focusing on highly relevant features. NB exhibited consistent accuracy across all feature sets.
Publisher
Engineering, Technology & Applied Science Research