Study of Approximate Periodicity Mining with Multi-granularity Time

This paper discusses a mining problem of approximate periodicity with multi-granularity time in the temporal database. It introduces the concepts and properties of the multi-granularity time interval on the basis of multi-granularity time and multi-granularity time format. It constructs multi-granularity approximate periodic pattern. It proposes an mining algorithm based on self-organizing map to find multi-granularity approximate periodic pattern. Results obtained from experiments on high frequency stock market data of 580000 Bao Steel JBT1 demonstrate that the proposed algorithm is efficient.