Evaluation of Risk Factors in Developing Breast Cancer with Expectation Maximization Algorithm in Data Mining Techniques

Abstract

Early detection and diagnosis of breast disease can improve the treatment effectiveness. Breast cancer is the second most common cancer for women. It is the second largest cause of cancer death worldwide. Annually, approximately more than 1,700,000 women worldwide are detected due to this disease. The prevalence of approximately 2 annual increased. Breast cancer involves several risk factors, some of which are proven but some still have controversial reported results and some are almost rejected. Sometimes factors such as maternal age at first birth, age at marriage and number of children have been recognized as risk factors, and sometimes as protective measures. In this paper we proposed a model that can predict the likelihood in developing a breast cancer. We modeled 7 different risk factors and their impact factors or their weighting using the data from Breast Cancer Surveillance Consortium (BCSC) in National Cancer Institute. We discovered the latent knowledge and generated new information by applying data mining techniques. Expectation Maximization (EM) algorithm was applied, data clustering was accomplished and the correlation of different risk factors was discovered. By analyzing our discovered information, we presented a novel formula to determine the probability in developing breast cancer and by using the proposed novel formula 98.6 accuracy was acquired.