Semi Automated Interactive Ensemble Based Learner

Abstract

In the real world as we know that there is no single best automated classifier for all the datasets, so there is this evident challenge of building intelligent interactive classifiers instead which can effectively work on a wide array of datasets. In the recent years many classification algorithms have been applied to diverse datasets for the purpose of classification. While successfully doing so with high accuracy these algorithms do not utilize the opinions of human experts on the data to the fullest extent. In this paper we have proposed a semi automated interactive approach to classification. Our algorithm utilizes expert opinions to classify data more effectively into different classes the user expects to visualize in that domain. Our proposed algorithm basically ensembles two classifiers to classify data with high accuracy without invoking the expert many times. It provides the user with an interface to allow him to edit or add a certain class for ambiguous records to incorporate sufficient domain knowledge and ultimately to suit the needs of the user's application.