Computational Intelligence and Healthcare Informatics

6Improving Multi-Label Classification in Prototype Selection Scenario

Himanshu Suyal* and Avtar Singh

Department of Computer Science and Engineering, Dr B R Ambedkar National Institute of Technology, Jalandhar, India

Abstract

ML-KNN is the one of the most common multi-label classifications, which is a modified version of the KNN. It is very simple yet powerful algorithm which always gives the promising results. Although it is very simple, the main hindrance of the classifier is the higher time complexity as well as sensitivity toward noise in the data set. To overcome this problem, prototype selection can be performed carefully. Prototype is the process to reduce the training data sets by selecting the most profitable prototype of the training data sets although prototype selections always compromise the accuracy of the algorithms. In this paper, a novel approach has been proposed to select the prototype without dropout the accuracy of the multi-label classification algorithm. To obtain the above objective, the well-known clustering algorithms are used to select the prototype by removing non-prominent data from the training data sets. The results show the improvement on ML-KNN.

Keywords: ML-KNN, prototype, multi-label classification, clustering

6.1 Introduction

Classification has widely played very important roles in many application fields and now became the hot research topic in the last decade. Classification is the process of assigning the incoming data into the predefine ...