A Comprehensive Analysis of Classification Methods for Big Data Stream

Traditional tools for mining the big data have become insufficient due to ever-growing data in the world. For handling big data, real-time and distributed processing is adopted. From so many mining tools options, it can be difficult for a researcher to opt an efficient tool. This paper is intended to aid the researcher who understands WEKA but is inexperienced with big data. The preliminary stage of data mining is classification, which categorizes the data into predefined groups. In this paper, WEKA with MOA package is used to classify big data stream with four different classifiers. Performance of these classifiers is analyzed on the basis of accuracy, i.e., correctly and incorrectly classified instances, time taken to test the model, and time taken to build the model. For this particular scenario, obtained results prove that naive Bayes is the most accurate classifier and decision stump is least effective classifier for big data.