Prostate Cancer Prognosis Using Multi-Layer Perceptron and Class Balancing Techniques

Abstract

Prostate malignancy is one of the most common malignancies. Early prediction of a cancer diagnosis can upsurge the endurance rate of cancer patients. The advancement of cancer research is boosted with the advent of artificial intelligence. Researchers have developed programmes to aid in cancer detection and prognosis due to the availability of open-source healthcare statistics. Machine Learning (ML) algorithms play a vital role in the field of cancer prognosis. The current study highlights the applications of neural networks to predict prostate cancer. We have accessed prostate cancer records from a publically accessible data repository (Kaggle). Current research work stresses the applications of neural learning approach for cancer prognosis and attaining more accurate prediction outcomes. The study also stresses on the impact of different balancing techniques on imbalanced data. The proposed method enhanced the accurateness from 72% on the imbalanced data to 97% on the oversampled dataset. This study aims to determine whether an artificial neural network (multilayer perceptron, MLP) can accurately predict the diagnosis of prostate cancer. In addition, the experimental results confirm the necessity of data balancing techniques in classification.