A fast and efficient algorithm for finding frequent items over data stream

We investigate the problem of finding the frequent items in a continuous data stream. We present an algorithm called λ-Count for computing frequency counts over a user specified threshold on a data stream. To emphasize the importance of the more recent data items, a fading factor λ is used. Our algorithm can detect ε-approximate frequent items of a data stream using O(log_λε) memory space and O(1) time to process each data record. The computation time for answering each query is O(log_λ ε), and for answering a query about the frequentness of a given data item is O(1). Experimental study shows that λ-Count outperforms other methods in terms of accuracy, memory requirement, and processing speed.