The Similarity Measurement Based On Constraint Dynamic Time Warping

Time series similarity matching is an important tool in data mining. Firstly, time series of different treatment technologies detailed summary of the basis of the previous analysis of existing mining tools; after the description of the existing problems, including time-series measurements and data stream processing, and Analysis of advantages and disadvantages of each method. Eventually made over for dynamic time warping matching problem, a detailed analysis of this issue based on the proposed application of the two algorithms in different fields, that Flex Dynamic Time Warping (FDTW) and Lining Bound Mine (LBM) algorithm, algorithm proof and experimental validation of their accuracy.FDTW DTW algorithm is mainly to solve the existing problems have made matching metric, which uses three iterations of the routing matrices strictly limited, if out of the path matching region cannot be the choice of points.LBM algorithm is applied in the time series data stream support for elastic deformation of the sub-sequence matching algorithm. As opposed to traditional static data flow time series are unique, cannot support the repetitive look, so the data flow sequence algorithm must be found through a scan area optimal solution. Algorithms have been proposed previous spring, but the existence of redundant computing algorithm and the spring had mismatch. LBM algorithm uses global constraints and determine the reduction of redundant computing over match and, after the experiment, the speed relative to the spring algorithm not only be enhanced, and will not match the phenomenon occurred.