Study On Asynchronous Periodic Patterns Mining In Time Series Based On Dynamic Link Structure

Periodic behavior existing in time series commonly, from the time series database to extract the periodic rule, that is, periodic pattern mining in time series plays an important role in data mining. Asynchronous periodic pattern mining is time series data mining frontier as a branch of a time series of periodic pattern mining. It also has important research value and practical application of theoretical significance. The current periodic pattern mining of the time series has focused on the synchronous periodic pattern mining, however, when there are some noise, some data sets missing, or specific data sets insert in the time series, periodic pattern misalign. The current synchronous periodic pattern mining is very difficult to find the hidden sequence periodic pattern. In the financial, transportation, electricity and bio-information time series, the asynchronous periodic pattern is widespread. There are a few of people are researching in this field. So this essay chooses asynchronous periodic pattern mining in time series as the main object of study. Firstly, the time series data mining and pattern mining cycle are reviewed, focusing on the progress of the mining asynchronous periodic patterns in time series, the basic definition and the current four asynchronous periodic pattern mining correlation algorithms of asynchronous periodic pattern mining is introduced in details. The four algorithms are two-phase algorithm, SMCA algorithm, OMMA algorithm and the E-MAP algorithm. All of this content is the basis of these elements, throughout the entire study process of periodic pattern mining in the time series algorithm.This article focus on the current four kinds of typical algorithms of asynchronous periodic pattern mining were compared from the mining object, process and results, scanning times of time series database, time complexity and space complexity on several aspects to analysis what the algorithm advantages and weaknesses of four algorithms. The direction of improvements in asynchronous periodic pattern mining can be found. Considering most of the time series is non-retroactive nature, an asynchronous periodic patterns mining algorithm is innovative proposed based on dynamic list structure. It uses the linked list structure and save storage space effectively. The time series database is scanned only once, the complex pattern needed by users will be getting without going through the two phases of the creation of single one pattern and multiple one pattern. This algorithm improves timeliness and effectiveness. The synthetic data and the real gene sequence data of the simulation incident the effectiveness of this article.