Solving Multi-Class Classification Tasks with Classifier Ensemble based on Clustering

Abstract

Ensemble learning is very popular for few decades for solving classification problems, because it generates and combines a diversity of classifiers using the same learning algorithm for the base-classifiers. In this paper we propose a method for generating classifier ensembles based on clustering. But with the continuous expansion of data availability in many large-scale, such as surveillance, security, Internet, and finance. It becomes critical to advance the fundamental understanding of knowledge discovery and analysis from raw data to support decision-making processes. Unremarkable computers can't effectuate the demand as they have unsubstantial memory space and delimited speed. As a result of these types of issues, contemporary prediction gets delayed. To eschew these problems, we have done some research and approached a efficacious algorithm so that we can reduce the amount of data by collecting only the informative data from the whole data set by using clustering. So that the mammoth data can be handled quite conveniently. After clustering we get new sub data sets. From each cluster we make fewer chunks of data than the archetype data set with the informative data with our new algorithm. After mingled these sub data sets we have run different ensemble algorithms on this new data set for comparability or exemplification. Comparing the results brandish that the accuracy rate is almost similar or increasing or decreasing. In some case we get more accuracy and in some case we get almost same accuracy. Once in a while we get less accuracy but even if we get less accuracy, the convenience side is that the amount of data is shortened.