Study On Theory And Algorithm Of Periodicity Mining In Temporal Data

Periodicity mining is the mining of periodic patterns, that is, the search for recurring patterns in time-series database. Periodicity mining can be applied to many important areas. For example seasons, daily traffic patterns, stock's fluctuating prices and trace of web access all present certain periodic patterns. First chapter describe the researching background and relative technology of data mining, analyze researching situation, discuss method and content of periodicity mining and sum up the result of this area. Focusing on method of periodicity mining, it educes the content of this paper. Second chapter we extend the strictly mathematic define of time, such as time type, time gene, time granularity. We present multiple granularities time format and prove some its property. Third chapter we researched two algorithms which mine partial periodicity patterns in time series. Algorithm CA utilizes the father first pruning method. Another presents a method that conversely computes the availability. Numerical experiment show the two algorithm offers good performance. Forth chapter we symbolize numerical time series by clustering its attributes. Then a asynchronous quick algorithm has been present for time series with noise event. The main contribution of this paper is presentation of multi-granularities temporal data and modal of partial periodicity. Get two algorithm which mine partial periodicity patterns in time series and one algorithm which mimes asynchronous periodicity in time series.