The Clustering Of Time-Series Data Based On LB_Hust Distance Caculation

Time series data is widely distributed, since a long time,lots of time series data in variety of areas was accumulated, the data mining on time series data gradually gets more attention. This form make public stock data time-series data as the object of studying, , the clustering of time series data as research purposes.by clustering of the stock data, we can make further analysis and research on the representation of similarity of the of time series.In this paper, for implementation of clustering based on trend similarity,we adopt the improved LB_Hust distance as the measure of time series similarity,which is a more suitable measure of the trend similarity of time series data,through by the partition of positive distance and negtive distance.and then implement time series clustering with the improved LB_Hust distance caculation.During implementation,we make some improvement to decrease complexity of the clustering algorithm.At last, we applicate the algorithm we proposed on the stock time-series data clustering,to find companies using similar operating methods and was influenced by the similar factors,so as to benefit long-term investors with decision-making guideThe experimental results show that the improved LB_Hust distance can be used to describe the similarity of time series better,and extend the range of application of the LB_Hust distance.