Support Vector Machine for Multiclass Handwritten Digits

In our research paper, we have implemented Multiclass Classification using Support Vector Machine (SVM). Pen Digit Recognition of Handwritten digit dataset is used for the purpose. One vs All approach has been applied using SVM to achieve multiclass classification. The same approach with different kernels has been analysed to select the right kernel. In this paper, we have found that selection of Kernel is very important in SVM classification process. The changes in kernel lead to variation in accuracy level from 11.2% to 97.68% which is quite significant. However, with the right selection of kernel the accuracy achieved with SVM classification is at highest level. Keywords—Support Vector Machine; Multiclass Classification; Kernel, Handwritten Digit Dataset.