Font Size: a A A

The Research Of Similarity Match Based On DTW In TSDM

Posted on:2011-04-24Degree:MasterType:Thesis
Country:ChinaCandidate:D C SunFull Text:PDF
GTID:2178360305978220Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Similarity Match is one of the hot researches in TSDM, which mainly includes three technologies: similarity distance, representation and seeking algorithm for Similarity Match. Similarity distance is used to measure the similarity between two time series. DTW is a kind of similarity distance in many similarity distances, which can effectively deal with part time-stress. But for the dynamic programming is directly used in this process, the time complexity is too high and the efficiency is low in practical application. Because time series is the massive and high-dimension dataset, which make the Similarity Match difficult finish. So the main nature in the original data must be manifested by a new representation and be transformed a better and succinct data. PLR is a kind of representation in many representations, which merits are such as effective Similarity Match, supporting new similarity distance,information feedback,text or data Series, new clustering algorithm, classification algorithm, singular points detection. Many improvement methods were created to improve the precision of the old PLR.Firstly, the present situation of Similarity Match was analyzed, and the research on technologies in similarity match was done. Then deep research on representation, especially on the PLR, was done. Through research,"PLR Based on Time Series Tendency Turning Point"was introduced. This is a method based on time series time-variable characteristic, aimed at extracting the tendency in the time series and compressing primary data, used the effective turning point to partition. This method is good at extracting the tendency in the series and compressing primary data, at the same time, it can partition time series as the series growing and has the merit of easily being carried out, remarkable result and suitable ability for the data from different field. After the existing DTW algorithm was analyzed, an improved efficiency algorithm was presented. Because the method avoided some repetition computation in the old algorithm, the improvement of the computing efficiency is obvious. Besides, the"PLR Based on Time Series Tendency Turning Point"and the"An Improvement Algorithm of DTW"were unified into"An Algorithm of DTW Based on Time Series Tendency Turning Point", which can accurately similarity match, at the same time, the improvement of the computing efficiency is obvious.
Keywords/Search Tags:Dynamic Time Warping, Time Series, Similarity, PLR, Tendency Turning Point
PDF Full Text Request
Related items