An amalgamated prediction model for breast cancer detection using fuzzy features

Abstract

Input feature processing is required for obtaining meaningful results for cancer prognosis. In this paper, the extended Kalman filter (EKF) and fuzzy K-means clustering algorithms have been combined into a hybrid algorithm with improved functionality, compared to either of the two separately. The proposed hybrid algorithm implements fuzzy K-means with support vector machine (SVM) coupled with an EKF for data filtering, working with from consecutive filtering and prediction cycles. Fuzzy membership functions are then calculated to map the labels with the attributes which is used by K-means to create a new modified set of attributes supplied to the SVM classifier, with lesser number of support vectors. The number of clusters is added into the training process as the input parameter except the kernel parameters and the SVM penalty factor. The approach was tested for various publicly available datasets like UCL, SEER and a real dataset compiled by the authors.