Optimizing Leukemia Classification Through Nature-Inspired Feature Reduction and Stacked Ensemble Learning Algorithms
Saumendra Kumar Mohapatra, Sharmila K.P, Mihir Narayan Mohanty · 2024
Leukemia, a complex and heterogeneous group of hematologic malignancies, presents significant challenges in accurate diagnosis and classification. In this work, an innovative approach to optimize leukemia classification using nature-inspired feature reduction approaches and stacked ensemble learning algorithms is proposed. Microarray data, a rich source of genetic information, often suffer from high dimensionality, making it challenging to extract meaningful features for classification. To address this, we employ nature-inspired algorithms such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Grey Wolf Optimizer (GWO), and Hunger Search Method (HSM) to reduce data dimensionality while preserving important features. Subsequently, we explore the efficacy neural network, including Deep Neural Networks (DNN) and stacked ensemble models, for classification tasks. Our test results show that, in comparison to conventional techniques, the suggested strategy considerably increases the accuracy of the leukaemia classification. Moreover, stacked ensemble models outperform individual classifiers, highlighting the efficiency of combining diverse models for improved classification performance. Overall, our study contributes to advancing precision medicine by providing a robust framework for optimizing leukemia classification through the combination of nature-inspired feature reduction and stacked ensemble learning algorithms.