A Novel Chaotic Kernel Framework for Support Vector Machines using Probability-Based Feature Extraction Method

This manuscript proposes a framework to reduce the uncertainty of the data points for chaotic kernel. The work applies a probability based feature extraction method to find the threshold value of the data point and determine the uncertain support vectors (SVs), The reduction of uncertainty in data points was achieved through the framework. The computational time was also optimized. The experimental details of the same are proposed henceforth.

I. Introduction

Classification is a technique to group the data points using some classification methods i.e. supervised and unsupervised [1]. Unsupervised classification is a task to group the data points on the basis of similar patterns without data labels, whereas supervised classification task is grouping of the data points with the corresponding data labels [2] [3]. In 1995, Vapnik proposed a new classification method to reduce the complexities of non-linear data using neural networks, called Support Vector Machine (SVM)[2] [4]–[7]. The Support Vector Machine (SVM) is a classification method used to classify the non -linear data in linear way with the help of kernel function [2]. The representation of support vectors is shown in fig. 1 below. The Support Vector Machine have been applied in many fields like text categorization, image classification, etc[8]–[10]. Fig. 1.

Support vector machine with adequate marginal space