An optimum approach to vector machine with K-nearest neighbor

Abstract

Kernel-based Virtual Machine (KVM) is an open-source computer technology built into Linux. The hybrid classification system is built by combining KNN and Support Vector Machine is abbreviated as K-VM. Due to its flexibility and reliability in handling different forms of classification activities, the KNN has been stated as one of the most frequently used classification approaches. The KNN faces a significant challenge in determining the acceptable value for parameter K to ensure good classification efficacy. This is because the value of parameter K has a significant effect on the KNN classifier’s accuracy. The KNN is a method of learning that is based on laziness that holds the entire training examples before classification time, in addition to deciding the optimum value of parameter K. As a result, as the value of parameter K increases, the KNN’s computational method becomes more intensive. To construct a new text message classifier, this paper combines the K-nearest neighbor (KNN)classification approach with the support vector machine (SVM) training algorithm. This paper proposes the K-VM hybrid classification system to reduce the impact of parameters on classification accuracy. The Euclidean distance function is used to measure the average distance between the testing data point and each range in SVs in various categories. Experiments on a variety of benchmark datasets show that the K-VM approach outperforms the conventional KNN classification model in classification accuracy.