Efficiently handling boundary frequent elements in stream data

Frequent pattern mining on stream data is a challenging task due to large volume and high velocity of stream data. As data are continuously evolving entire dataset cannot be scanned. So data under a particular window are scanned for finding frequent patterns. Among different window models available, sliding window is the most popular as it focuses more on recent data. However sliding window model suffers with a problem of false negatives i.e. some items which are frequent are shown as infrequent. This problem occurs as boundary frequent elements are not detected. The novel algorithm proposed EHBFS efficiently handles this problem and finds frequent patterns on stream data with more accuracy. Extensive experiments have shown that frequent patterns produced are better than variable size algorithm and MOA Incmine.