Ovarian cancer substantial risk factor analysis by machine learning: A low incoming country perspective

Abstract

In this paper, ovarian cancer data were inspected to figure out the significant risk factors. According to the American Cancer Society, it is the fifth leading cause of death of women. It characterizes that in 2019, 22,530 women will be diagnosed, whereas 22,240 women were diagnosed in 2018, and 13,980 will face death in 2019, but 14,070 women were died because of ovarian cancer. For this research, 521 woman’s data were collected from Hospitals of Dhaka with case group 267 and control group 254. A set of a questionnaire of 47 factors that were elicited from various researches used for data collection. Data were examined with different machine learning algorithms like using SVM, logistics regression, random forest, naïve bayes, neural network, kNN, ada boost, CN2 rule, Decision tree, Quadratic Classifier. These algorithms were compared with each other with different tools and found that Logistics Regression provides the highest accuracy of 0.933 along with the highest CA of 0.848. Data were investigated with ranker algorithms to found out the rankings between factors with the help of feature selection. Significant factors like problems during pregnancy, abortion, cervical cancer history, menopause problems, etc., were found out as significant risk factors of ovarian cancer.