Efficiently mining frequent patterns in recent music query streams

Abstract

Mining frequent melody structures from music data is one of the most important issues in multimedia data mining. In this paper, we proposed an efficient online algorithm, called BVMDS (bit-vector based mining of data streams), to mine all frequent temporal patterns over sliding windows of music melody sequence streams. An effective bit-sequence representation is used in BVMDS to reduce the time and memory needed to slide the windows. An effective list structure is used to overcome the performance bottleneck of previous work, FTP-stream. Experiments show that the BVMDS algorithm outperforms FTP-stream algorithm, and just scans the streaming data once.