Mining discriminative items in multiple data streams with hierarchical counters approach

In this paper a 1-pass algorithm is presented for finding the discriminative items between multiple data streams using very limited storage space. The approach relies on a novel data structure called hierarchical counters. The number of each item is shown by one of the 10-valued positions in the counters, which allows us to identify the exact frequencies of all the items in the streams and also discriminative items between multiple data streams based on user identified parameters. In contrast with previous works, this approach is not limited to user predefined parameters and discriminative items could be identified based on different thresholds, without needing any change.