Collective Sequential Pattern Mining in Distributed Evolving Data Streams

The advances in processing and communication techniques resulted in a multitude of emerging applications that interact with streams of data. Traditional data mining systems store arriving data, collect them for later mining, and make multiple passes over the collected data. Unfortunately, these systems are prohibitively slow when they deal with data streams with massive amounts of data arriving at high rates. This paper introduces a new model for mining sequential patterns on distributed data streams environments. It focuses on evolving data streams that originate from multiple distributed sources. Moreover, the mining process is achieved without compromising the privacy of the individual data streams of the participant nodes. Simulation results show that the proposed model scales linearly with the number of distributed nodes. In addition, it reduces the overhead in the distributed mining process.