Prediction of Bladder Cancer Recurrence Using Classification Methods
Donghyok Suh, Dong-Mun Shin, Ho Sun Shon, Wun‐Jae Kim, Won-Tae Kim, Keun-Ho Ryu · 2012
Bladder cancer is a malignant disease that occurs in the urinary bladder. It is the most common type of cancer after prostate cancer in the whole world. Bladder cancer can occur at any age, but it is more common among older people than the younger ones. In this paper, we built several classification models to predict the recurrence of bladder cancer with various well-known feature selection and classification techniques. In our experiment, we selected the most suitable data mining model to predict the risk of bladder cancer recurrence using several classification algorithms, such as artificial neural network, bayesian network and support vector machine. Our experiments proved that minimum redundancy maximum relevance criterion is better than conditional mutual information maximization, especially, the highest classification accuracy is obtained by bayesian network model, which contains only 10 important features. Therefore, we derived a conclusion that bayesian network model is the most appropriate for predicting the risk of bladder cancer recurrence, when comparing with other classification methods. We expect to give help in preventing cancer and estimating prognosis by discovering useful knowledge that is extracted from the model.