An Optimised Hybrid Approach for Multi-Class Cancer Classification in Microarray Data

Nearly nine million people worldwide suffer from cancer, according to the WHO. Machine learning (ML) has the potential to diagnose cancer in genetic data, such as microarray gene expression patterns, thereby reducing mortality rates. But high-dimensional datasets sometimes limit model performance. Mitigating this issue necessitates the implementation of feature selection algorithms to streamline data while preserving essential information. Ant lion optimization (ALO) and ant colony optimization (ACO) feature selection and SVM classification are used for multi-class cancer classification in this research. When assessed on the ALL-AML and Lymphoma microarray datasets, the approach demonstrates exceptional classification accuracies of 95.22% and 94.77%, respectively, demonstrating its potential for effective cancer diagnosis.