An Improved LB_Hust Distance Calculation Method
LB_Hust distance calculation method developed from DTW, it is accumulativeness of unsigned, no weight points distance: tbe calculating result is the accumulation of the absolute value of all the points at even- time in the time series, no considering the time sequence trend change influence. So the improvement of the traditional LBHusI distance and separation of the time series data considering the influence of the trend of time series is researched in this paper. The form for calculating the distance called LBHust symbols distance calculation method is obtained. And the practical stock data is used as database, the original LBHust distance calculation method and LBHust symbols distance calculation are realized on distance matrix of hierarchical clustering algorithm. Many-stocks are clustered according to the similarities of trend. The results show that the improved LB_Hust symbols distance calculation method in similarity calculation can describe the trend of time sequence comparability more accurately, the same trend of the time sequences are gathered to a class, the application scope of LBHust distance calculation method is expanded.
time series DTW algorithm LB_Hust distance LB_Hust symbols distance
Liu Jiang Gong Wenqin Cui Meiling
School of Computer Science and Technology Tianjin University Tianjin, China
国际会议
海口
英文
305-309
2011-07-15(万方平台首次上网日期,不代表论文的发表时间)