Efficient Frequent Pattern Mining in Data Streams

Abstract

Listen

Translate article

In this paper, we present a new frequent pattern mining mechanism for data streams to enhance the performance of existing mechanisms, particularly to reduce the needed processing time and uplift the mining accuracy. Our new mechanism performs data mining by the Trie data structure (instead of the entry table in previous mechanisms) to access patterns at reduced processing steps and run time. To attain desirable mining accuracy, it adopts a Length Skip practice to skip less frequently used long patterns from the summary and preserve the space for more frequently used short patterns. Simulation results show that, when incorporated into existing frequent pattern mining algorithms, our new mechanism will effectively increase the mining accuracy at notably reduced run time.