Robust Classification using Twosome Classifier

Abstract

The proficient knowledge revolution with the concept class detection is a significant challenge for the incremental learning classification, here the batch of data is flowing unremittingly. The algorithm recommended in this research learns continual raw data, transform preceding knowledge to the present data without stating to the longstanding data and able to proficiently allow new concept class noticed by the classification. A major aim was knowledge transformation and the accumulation of the same with concept class detection in the incremental data flow. After the analysis of non-incremental ML approaches for classification and incremental learning algorithms in the literature, this research introduces new incremental learning algorithm, which uses twosome classifiers, in which knowledge transformation and new class detection have been done efficiently. The performance of the system is validated using simulation results on available datasets. The class wise and batch wise accuracy is calculated. The projected technique is also used to detect concept class detection.