Clustering and Nearest Neighbour Based Classification Approach for Mobile Activity Recognition

Abstract

We present a hybridized algorithm based on clustering and nearest neighbour classifier for mobile activity recognition. The algorithm transforms a training dataset into a more compact and reduced representative set that lessens the computational cost on mobile devices. This is achieved by applying clustering on the original dataset with the concept of percentage data retention to direct the operation. After clustering, we extract three reduced and transformed representation of the original dataset to serve as the reference data for nearest neighbour classification. These reduced representative sets can be used for classifying new instances using the nearest neighbour algorithm step on the mobile phone. Experimental evaluation of our proposed approach using real mobile activity recognition dataset shows improved result over the basic KNN algorithm that uses all the training dataset