Breast Cancer Classification by Gene Expression Analysis using Hybrid Feature Selection and Hyper-heuristic Adaptive Universum Support Vector Machine

Comprehensive assessments of the molecular characteristics breast cancer from gene expression patterns can aid in early identification and treatment tumor patients. The enormous scale data obtained through microarray sequencing increases difficulty training classifier due to large-scale features. Selecting pivotal features minimize high dimensionality complexity with improved detection accuracy. However, traditional filter wrapper-based selection methods have scalability adaptability issues handling complex This paper presents a hybrid feature method Mutual Information Maximization - Improved Moth Flame Optimization (MIM-IMFO) for along an advanced Hyper-heuristic Adaptive Universum Support classification model Vector Machine (HH-AUSVM) improve rates. is developed by performing filter-based using MIM first stage followed wrapper second stage, obtain remove inappropriate ones. improves standard MFO exploration/exploitation phase accomplish better trade-off between exploration exploitation phases. HH-AUSVM formulated integrating learning approach hyper- heuristics-based parameter optimized SVM tackle class samples imbalance problem. Evaluated on datasets Mendeley Data Repository, this proposed MIM-IMFO selection-based provided accuracies 95.67%, 96.52%, 97.97% 95.5% less processing time 4.28, 3.17, 9.45 6.31 seconds, respectively.