Extraction of relevant dataset for support vector machine training: A comparison

Support Vector Machine (SVM) is a popular machine learning technique for classification. SVM is computationally infeasible with large dataset due to its large training time. In this paper we compare three different methods for training time reduction of SVM. Different combination of Decision Tree (DT), Fisher Linear Discriminant (FLD), QR Decomposition (QRD) and Modified Fisher Linear Discriminant (MFLD) makes reduced dataset for SVM training. Experimental results indicates that SVM with QRD and MFLD have good classification accuracy with significantly smaller training time.