Optimised deep neural network for cancer disease prediction using a genetic algorithm

Abstract

Detection of cancer at an early stage is a crucial activity for the oncologist for proper treatment of the disease. Various machine learning techniques are applied to detect different types of cancers. However, till date, the low accuracy in cancer detection is observed because limited focus has been given to addressing the dataset imbalance problem for cancer detection. In the present work, a novel genetic algorithm-based deep neural network (GA-DNN) is proposed to effectively detect the two types of cancers i.e., prostate cancer and breast cancer. Results obtained by GA-DNN are compared with support vector machine (SVM), random forest, and deep neural network (DNN). Best results are reported by GA-DNN for prostate cancer and breast cancer Coimbra datasets. It was observed that the optimised DNN gave the best results when the dataset is large.