Cluster time series based on partial information

Cluster analysis has been studied extensively for many years. Recently, there has been a lot of interest in using cluster techniques in mining time series data. The previous cluster models mainly focus on the overall and the most remarkable series behaviors. We address the problem: using a portion of information in clustering time series. We describe a model for retrieving and representing the partial information in time series data. By using this model, a methodology for partial-information-based clustering is proposed. We evaluate our approach through comparing the results with a standard classification. The results show our approach could outperform the previous clustering method that is based on the whole information of time series.