ABSTRACT
This chapter discusses proposed framework for classifying data streams with automatic novel class detection mechanism. Existing ensemble techniques in classifying concept-drifting data streams follow a single-partition, single-chunk approach in which a single data chunk is used to train one classifier. Existing data stream classification techniques assume that the total number of classes in the stream is fixed. ECSMiner outperforms the state-of-the art data stream-based classification techniques in both classification accuracy and processing speed. ECSMiner is applicable to the more realistic scenario where there is more than one existing class in the stream. Realistic Data Stream Classifier (ReaSC) is a more practical approach to the stream classification problem since it requires a lower amount of labeled data, saving much time and cost that would be otherwise required to manually label the data. Real-time data stream classification is important in many applications such as the intrusion detection systems.
