Using Bayesian networks and importance measures to indentify tumour markers for breast cancer

Because breast cancer has become one the most common cancer among women, this paper identified some effective tumour markers from historical patient records to support cancer diagnosis. First, the advantages of Bayesian network (BN) in target classification are discussed, and the concept of importance measures are introduced. Then, the original breast cancer data records used for case study are collected from the first affiliated hospital of medical college of Xi'an Jiaotong University, China, which are also discretized and cleared to form the standard modelling dataset. Finally, the practical BN model of each target variable is learned from the dataset respectively according to the tumour marker variables of breast cancer. Based on the constructed BN models, the importance values of all tumour marker states are calculated and discussed for tumour marker identification.