Title: Emsembling imbalanced-spatial-structured support vector machine
Authors: Xin Liu - Shanghai University of Finance and Economics (China) [presenting]
Abstract: The Support Vector Machine (SVM) and its extensions have been widely used in various areas due to its great prediction capability. However, these methods cannot effectively handle imbalanced data with spatial association which commonly arises from many studies such as cancer imaging study. We propose the ensembling imbalanced-spatial-structured support vector machine (EISS-SVM) method, useful for both balanced and imbalanced data. Not only does the proposed method accommodate the association between the response and the covariates, but also accounts for the spatial correlation existing in the data. Our EISS-SVM classifier embraces the usual SVM as a special case. The proposed method outperforms the competing classifiers shown in both the simulation studies and an application to real imaging data from an ongoing prostate cancer research conducted in Canada.