Improved Hierarchical Clustering Using Time Series Data

Abstract

Mining Time series data has a remarkable development of interest in today’s world. This paper presents and evaluates an incremental clustering structure for time series data stream. The new algorithm is called Improved Hierarchical Clustering Algorithm (IHCA) is developed and applied with ECG data set. This system continuously constructs a tree structure of hierarchy that progress with data set. Two kinds of operations need to grow the Hierarchical clustering algorithm. The operations are split and merge (reaggregate). According to the diameter of the cluster the specific operation is decided. The split operation is based on dissimilarity measure between time series data points. The merge operation is to combine a previous split node in order to reacts the changes in the correlation structure between time series data points. These two operators are adopting the fast arrival of time series data flow. Cluster quality, Outlier and compilation time are the main features of this research. Experimental results shows that the performance of cluster quality and computation speed are improved.