Performance Analysis of Classifier Techniques at the Edge Node

Abstract

A smart home contains smart gadgets which are generating a huge amount of data. The utilization of IoT gadgets in smart homes is constantly expanding and for faster processing of data, appropriate resources will be required. This generated data is passed to the cloud for processing but there may be a delay in the processing of data. Edge devices reside at the edges of smart gadgets and perform quick processing of data. Computation speed can be increased if generated data is classified and assigned to the edge node. The classifier is commonly used in machine learning algorithms. It can also be used in smart city and smart home applications. Data classification helps in decision-making by finding outliers from data. Many algorithms are available for data classification and out of which the rule-based classifier [1] and k-means clustering [2] are the most commonly used classifier. In this paper, we attempted to analyze the performance of the rule-based classifier and k-means clustering based on evaluation parameters viz. average execution time, service latency, and resource utilization. From the simulation results, it is observed that k-means clustering performs better as compared to rule-based classifier.