Abstract
Time-series clustering algorithms have been used in a variety of areas to extract valuable information from complex and massive data sets. However, these algorithms suffer from two shortcomings. On the one hand, most of them are designed for the equal-length time series, while clustering of unequal-length time series is often encountered in real-world problems. On the other hand, commonly used distance measures of time series cannot fully reveal trend differences. To overcome these two shortcomings, this paper focuses on the trend of time series and employs the area-based shape distance to measure their similarity. In addition, we present a new hierarchical clustering for unequal-length time series based on area-based shape distance measure. A series of experiments illustrates the performance of the proposed clustering algorithm.