Frequent Items Mining Based on Weight in Data Stream

Frequent items mining is a very basic but important task in the data stream processing. However the traditional algorithms such as Lossy Count can only find out frequent items based on computing their counts. In some situations, people want to monitor those items whose weight exceeding a user-specified threshold over the data stream. In this paper, we propose a novel algorithm to address this problem. The Lossy Weight Algorithm can output an approximate result whose error is guaranteed not to exceed a user-specified parameter. Experimental results show that the new algorithm yields very good performance on both space and time cost. We believe that no previous work on weight-based frequent items mining exists.