Comparative Analysis of Classification based Techniques using Change Detection and Ensemble models

Abstract

In the data streaming environment, more and more data gets generated due to various applications like network intrusion detection, weather forecast, finance, sensor based data acquisition systems and so on. Due to rapid advancement in technology, the arriving pattern is expanding and hence data arrives continuously which need to be monitored rather than storage. Some of the application requires a quick analysis of arriving data. In such situations, storing such continuous and huge data and then analyzing the performance will not be useful for better decision.  Since the probability distribution of such data changes over the period of time it is necessary to timely analyze such patterns in the data such change is occurred due to unstable data distribution. This raises a need of continuous monitoring of such changes in in streaming environment the researchers have given major attention to concept change which is referred as concept drift. In this paper we have surveyed the various classification-based methods for handling change in the distribution, in introduction   section we have discussed various types of concept drift and their application, followed by the methodologies used to handle the diversity and further, discussed the experimental work performed on Covid-19 dataset. The last section concludes how the concept drift learners are extensively used in covid-19 data to detect the change in the distribution and how an adaptive learners and evaluators are   used for solving real life applications.