Comparative performance analysis of machine learning classifiers on ovarian cancer dataset

Abstract

Machine learning classifiers help physicians to make near-perfect diagnoses, minimizing costs and time. Since medical data usually contains a high degree of uncertainty and ambiguity, proper ordering and classification require a proper comparative performance analysis of machine learning classifiers. Machine learning classifiers are applied on the Ovarian Cancer Dataset. Ovarian cancer is the fifth leading cause of cancer-related death among women, and is the deadliest of gynecological cancers. The mortality rate of ovarian cancer ranks first. Thus, early diagnosis and treatment are critical for improving the patients' cure rate and prolonging their survival. Here we have investigated Mass spectrometry (MS) field data to develop a computer-aided system for the purpose. Using machine learning techniques, data is classified in different categories to identify benign and malignant cancerous cells and a comparative study has been done to identify the most suitable technique under different operational conditions and datasets. Our Comparative studies show that the Multilayer Perceptron (MLP) is the best options for such detection considering its performance metrics such as Accuracy, Sensitivity, Specificity and Errors