Modified Hierarchical Clustering Algorithm for Time Series Data

摘要

Time-Series clustering is used to attain deep knowledge of the mechanism that generate the time-series and speculate the prospective values of the given time-series. Time-Series clustering is shape-level if it is carried out on the many individual time-series or structure-level if it works on single long-length time-series. Depending on whether Time-series clustering is working directly on unprocessed data (frequency or time domain), or indirectly with the features extracted or model built from the unprocessed data, it is categorized into three groups. The proposed work comes under the raw data based approach. In this work, DTW is utilized as a distance/similarity count in the hierarchical clustering algorithm with inter/intra-cluster-distance-based-swap. The performance of the proposed work is evaluated by using Clustering Validity indices.