ReliefF and Association Rule Mining to Determine Cervical Cancer Causes

Abstract

Cancer patients till this day suffer from the inability of science to predict the causes of the disease before it occurs. One of the cancers that occupy the minds of many women is the cervical cancer because of the delay in its diagnosis as a result of its multiple and unclear causes, so scientists and researchers need to search for the most causative factors. Machine learning approaches have become one of the best and fastest ways to find associations between symptoms and causes of disease. The use of association rule mining (AR) is very effective if diagnostic features are set up. In this work, feature selection (FS) algorithm named ReliefF is used to reach the most correlated factor, then the Apriori algorithm has been updated to reduce the time and space used, and detects features that are closely related to the class attribute to access most factors that cause cervical cancer. The experimental results of the proposed work indicate a number of cervical cancer risk factors that when combined, indicate a woman's likelihood of developing cervical cancer, which is: the number of years of hormonal contraception is greater than or equal to 15, having any type of cancer or HPV or syphilis or HIV, the number of IUD insertion years exceeded 10, First sexual intercourse smaller than 13 and Number of sexual partners greater than 5. The outcomes of this work help both doctors and women to prevent cancer.