Development and Investigation of Data Stream Classifier Using Lagrangian Interpolation Method

Abstract

Analyzing data streams floated over digitized organizations have become a notable challenge as it is often being obstructed by the incidence of concept drift and concept evolution. Conventional classification algorithms are neither accurate nor reliable on classifying dynamic data streams. For instance, Support Vector Machine (SVM) is a prominent classifier for performing supervised classification on static data; however, it becomes flabby when it is directly applied for data stream classification due to its intrinsic nature. In a bid to avoid this issue, this research work expounds a novel SVM-based Parallel Genetic Ensemble Model, which makes a series of optimization on SVM by synergizing it with Lagrangian interpolation method, fuzzy logic, and parallel genetic algorithm. The proposed model effectively classifies real-time data streams amid of concept drift and concept evolution. The exhaustive experiments probed on real-time data streams assert the efficacy of the proposed system in terms of various metrics.