Comparison of Machine Learning Algorithms for Cervical Abnormality Detection

Cervical cancer occurs in the cells of the cervix which is the lower part of the uterus that connects to the vagina. According to WHO data, globally cervical cancer is fourth most prevalent form of cancer affecting women. 85% of deaths due to cervical cancer occur in low resource countries. A large number of these deaths occur due to non-availability of specialized doctors to conduct cervical screenings. Additionally, a number of times the hospitals in such areas are not equipped with medical facilities that are required to carry out cervical screening tests like pap smear, colposcopy and biopsy. Hence an automated system which can predict the risk of cervical abnormality can be very useful for speedy detection and treatment. In this work, we present Machine Learning(ML) methods which use demographic data rather than image data to detect whether a patient has cervical abnormality. Furthermore, a comparison of the performance of 10 different ML techniques has been presented.