A performance assessment of Bayesian networks as a predictor of breast cancer survival

摘要: This paper is intended to assess the survival analysis of Bayesian Network models, Neural and Logistic Regression models. Our will be performed on SEER data set, a registry women with breast cancer from National Cancer Institute. Each model include following prognostic variables; progesterone (PR), estrogen (ER), lymph involvement (N), morphology (M), extension tumor size (T), histological grade (G). These variables have proven significant in regards survival. We found that network model, which combination an automatically generated by BKD software human expert knowledge, performs comparatively better than Networks logistic regression The also offers advantage explaining causal relationships among variables, thus it most promising