Optimal Data Mining Method for Predicting Breast Cancer Survivability

Breast cancer is one of leading causes of death. This study predicts 5-year survivability of breast cancer patients by two data mining techniques. The data set consisted of information about patients who have cancer diagnosis collected by SEER. In this study, data set is pre-classified into survival and non-survival with 90.66% and 9.34%, respectively. The selected variables used to predict 5-year survivability of breast cancer patients are race, grade, extension of disease, site-specific surgery code, stage of cancer, and SEER modified AJCC stage 3. The performances of two methods are considered from the perspective of three criteria (i.e. accuracy, g – mean and ROC); the results show that logistic regression model is better than decision tree model.