Multi-objective hyperparameter optimization on gradient-boosting for breast cancer detection

The most commonly occurring cancer among women, breast cancer, causes lakhs of deaths annually, which can be prevented by early detection and treatment. Detection done using machine learning models on histopathological images are affordable, reliable, accurate. Previous studies in this regard have focused transfer methods combining feature selection Convolutional Neural Networks (CNNs) an ensemble gradient-boosting algorithms. However, none the state-of-the-art techniques capture multi-objective nature Breast Cancer (BCD) tend to improve a single performance measure such as Accuracy F1 score, fail certain essential aspects problem cost misclassification varies greatly depending its type. In study, hyperparameter optimization technique for Prediction is proposed comparing random search, Non-Dominated Sorting Genetic Algorithm (NSGA-II) Bayesian optimization. This approach applied three popular techniques: extreme gradient-boosting, light categorical boosting features obtained from Inception-ResNet-v2 CNN model benchmark BreakHis dataset optimize Precision, Recall, Accuracy, AUC simultaneously. novel NSGA2-IRv2-CXL study achieves maximum 94.40%, 98.16, Precision 95.77%, Recall 99.29% 100 $$\times$$ magnification. also establishes trade-offs between metrics thereby opening avenues further research approaches BCD provide larger view strengths weaknesses classification model.