A Comparative study of Data stream classification using Decision tree and Novel class Detection Techniques

The rapid development in the e-commerce and distributed computing generates millions of the transaction, continuously. This continues arrival of data is considered as a DataStream. Data mining process for classification needs considerable modification to cope with continuous data. As Mining continues stream of data, conceptually has infinite length, and the class of data may change in sudden or gradual or hike, for which classification model is completely unknown or not prepared. Here investigation is made on different techniques proposed for the data stream classification using decision trees. Different approaches of decision tree classification for the stream data are analyzed & compared. The primary comparison parameters are time and accuracy. Also shown efforts made for handling the change in the concept and they are compared in terms of memory, technique and accuracy. Index Terms Data stream, Novel class, Incremental learning, Ensemble Technique, Decision tree, Concept drift ___________________________________________________________________________________________________