Error-Adaptive and Time-Aware Maintenance of Frequency Counts over Data Streams

Maintaining frequency counts for items over data stream has a wide range of applications such as web advertisement fraud detection. Study of this problem has attracted great attention from both researchers and practitioners. Many algorithms have been proposed. In this paper, we propose a new method, error-adaptive pruning method, to maintain frequency more accurately. We also propose a method called fractionization to record time information together with the frequency information. Using these two methods, we design three algorithms for finding frequent items and top-k frequent items. Experimental results show these methods are effective in terms of improving the maintenance accuracy.