[[alternative]]Mining Repeating and Sequential Patterns from Music Databases

[[abstract]]In this paper, two research issues on music data mining are studied. The first one is mining repeating patterns in a single music and the other one is mining common sequential patterns in a set of music. The common sequential patterns can be classified further into two kinds of patterns: one is large sequential pattern and another is non-continuous sequential pattern. These kinds of patterns can represent the important features of music, which can be used for music data retrieval or music classification. The MRP algorithm is proposed for mining repeating patterns. The basic idea is to add one note at a time to produce the candidate note streams. Then the bit index stream representation is designed for counting the appearing frequency of a note stream efficiently in order to verify the repeating patterns. In addition, the appearance bit stream representation is proposed to cooperate with the bit index stream representation for mining the large sequential patterns. Moreover, by applying the pattern join algorithm, the large non-continuous sequential patterns will be found from the combinations of the large sequential patterns. Finally, the result of the experiments shows that the proposed algorithms have better performance than the related works.