Breast Cancer Prediction Using Hybrid Machine Learning and Nature-Inspired Adaptive Optimization Algorithms

Abstract

An early stage of breast cancer occurrence does not have pain as a symptom. This asymptomatic nature of cancer necessitates the need of timely and accurate prediction using other potential indicators. Breast cancers are highly curable if predicted and diagnosed at the earliest. This research explores the capacity of hybridizing intelligent learning models with nature-inspired optimization algorithms for breast cancer prediction. This integration becomes pivotal to optimize internal parameters of the dataset while aiming for augmented accuracy of the proposed predictive models. Each of the four intelligent machine learning models including K-Nearest Neighbor (KNN), Naïve Bayes (NB), Support Vector Machine (SVM), and Artificial Neural Network (ANN) are hybridized with all the three adaptive optimization approaches including Adaptive Particle Swarm Optimization (APSO), Adaptive Genetic Algorithm (AGA) and Adaptive Venus Flytrap Optimization (AVFO). Thus twelve hybridized predictive models were derived namely APSO-KNN, AGA-KNN, AVFO-KNN, APSO-NB, AGA-NB, AVFO-NB, APSO-SVM, AGA-SVM, AVFO-SVM, APSO-ANN, AGA-ANN and AVFO-ANN. These hybridized models were investigated through UCI data repository’s breast cancer dataset namely WBC, WDBC and WPBC. The experimental results were validated against the respective learning models with and without feature selection. Based on the performance measures such as accuracy, F-measure and G-mean, the outperformed learning models with WBC, WPBC and WDBC datasets are AVFO-SVM, AVFO-ANN and APSO-ANN respectively. Conclusively ANN and SVM machine learning algorithms fused with AVFO for feature selection are robust for all the three datasets. The derived hybrid intelligent models trained with tuned datasets optimize the prediction ability of existing breast cancer prediction models.