Finding Heavily-Weighted Features in Data Streams