Research On Time Series Data Mining Algorithm Of Partial Periodic Patterns

As new field of Time Series Data Mining, it is much valuable to research on data mining of periodic pattern.However, The mining of partial periodic patterns and Incremental mining is hard and important about it,so this paper chooses the mining of partial periodic patterns as the main task to study.First,this paper summarizes status of the Time Series Data Mining and Partial Periodic Patterns Mining in time series.then,It analyses the conception of time series,period,partial periodic patterns,L-length of pattern and Incremental time series.after that,the Aprior-like on periodicity,Max-subpattern Hit-set Algorithm and Entire-Segments Incremental Algorithm(ES) is studied.They are the foundation of research on Algorithm of Mining Partial Periodic Patterns and they can be used in the whole paper.Reviewing the Max-subpattern hit-set Algorithm and its character,this paper comes up a chained data structure to improve it.then it compares the time complex with the Max-subpattern hit-set Algorithm.After that,we research on Incremental mining in time series with chained data structure,based on Entire-Segments Incremental Algorithm.It keep much advantage of saving and separating from chained data structure,but it still has much limit.At last, we compare Incremental mining based on chained data structure with none Incremental mining ,at the same time,we also anylise the efficiency when confidence level changes.The innovation is that we come up a chained data structure to maintain Hit-set instead of Max-subpattern hit-set Tree.Chained data structure keep Hit-set by its L-length.it does not need to create ancestor nodes,so it can reduce the saving of none hit-set and the room of matching the patterns. At other hand, Chained data structure can be used in Incremental mining. It is not much efficencier than the tree structure,but it still keep the advantage of saving and separating the frequent patterns.