Bayesian Networks, Rule Induction and Logistic Regression in the prediction of the survival of women suffering from breast cancer
Pedro Larrañaga, Miren J. Gallego, Basilio Sierra, L. Urkola, Maria J. Michelena · 2000
In this paper we present an empirical comparison among several paradigms coming from Statistics and Articial Intelligence for solving a supervised classication problem. The empirically compared paradigms are Bayesian Networks, Rule Induction and Logistic Regression. The problem to tackle is the prediction of the survival of women diagnosed as having breast cancer taking into account four predictor variables gathered at the moment of diagnosis. The data le includes 1000 cases diagnosed at the Oncological Institute of Gipuzkoa (Basque Country). The validation of the paradigms was carried out using the 10-fold cross-validation method.