Classification of breast cancer microarray data using Radial Basis Function Network

Breast cancer is the number one killer disease among women worldwide. Although this disease may affect women and men but the rate of incidence and number of deaths are high among women compared to men. Early detection of breast cancer helps to increase the chance of survival since early treatment can be decided for the patients who suffer from this disease. The advent of the Microarray Technology enables it to be applied to the medical area in terms of classification of cancer and diseases. By using the microarray, the expressions of thousands of genes can be determined simultaneously. However, this microarray suffers several drawbacks such as high dimensionality and contains irrelevant genes. Therefore, various techniques of feature selection have been developed in order to reduce the dimensionality of the microarray and also to select only the appropriate genes. For this study, the microarray breast cancer data, which is obtained from the Center for Computational Intelligence, is used in the experiment. The Relief-F Algorithm is chosen as the method of the feature selection. As for comparison, another two methods of feature selection, which are Information Gain and Chi-Square, are also used in the experiment. The Radial Basis Function (RBF) Network is used as the classifier to distinguish between the cancerous and noncancerous cells. The accuracy of the classification is evaluated by using the chosen metric, namely Receiver Operating Characteristic (ROC).