Kaal - A Real Time Stream Mining Algorithm
Rajanish Dass, Varun Kumar · 2010
Finding frequent patterns in a data stream has been one of the daunting tasks since its inception. Mining data streams are allowed only one look at the data, and techniques have to keep pace with the arrival of new data. Furthermore, dynamic data streams pose new challenges, because their underlying distribution might be changing. Most importantly, the stream mining algorithm must be fast enough to adapt itself to slow as well as very fast data streams. In this paper, we have introduced a new stream mining algorithm called Kaal - Sanskrit word for time - that is significantly better than existing classical algorithms. Further, Kaal is capable of adapting well to variable batch sizes. The batches are decided by a fixed time quanta, any number of transactions coming in that time interval constitutes that batch. Previous stream mining algorithms demand fixed batch sizes, which in real world scenario becomes difficult to realize or fail to provide periodic real-time results.