Hybrid classification procedure using SVM with LR on two distinctive datasets

Traditionally, Support Vector Machines (SVMs) are used in classification and pattern recognition, which is also the case in Neural Networks and other learning algorithms. However, a major limitation is that when the training sets are imbalanced, SVM cannot perform satisfying results by the chosen kernel. To overcome this limitation, we propose a hybrid method which uses SVM linear kernel with Logistic Regression (LR) in different manner. The hybrid method is tested over two datasets. The results showed that the hybrid approach enables us to develop an efficient algorithm, which solves the problem with all imbalanced dataset training at one time. At the same time, when it is compared with SVM that uses Radial Basis Functions (RBF) and linear kernel, better accuracy estimations are achieved with promising results on classification compared with SVM that uses RBF and linear kernel.