An Adaptive Density-Based Time Series Clustering Algorithm: A Case Study on Rainfall Patterns

Time series clustering algorithms have been widely used to mine the clustering distribution characteristics of real phenomena. However, these algorithms have several limitations to mine clustering characteristics in geography. First, current time series clustering algorithms fail to effectively mine clustering distribution characteristics of time series data without sufficient prior knowledge. Second, the algorithms ignore the spatial heterogeneity of geographical objects. Thirdly, these algorithms fail to simultaneously consider non-spatial time series attribute values and non-spatial time series attribute trends, which are all important similarity measurements. In view of these shortcomings, an adaptive density-based time series clustering (DTSC) algorithm has been proposed in this paper. DTSC algorithm simultaneously considers the spatial attributes, non-spatial time series attribute values, and nonspatial time series attribute trends. DTSC algorithm proceeds with two major parts. In the first part, the objects with spatial proximity relationship are considered as similar in the spatial domain. In the second part, an improved density-based clustering strategy is then adopted to detect clusters with similar non-spatial time series attribute values and time series attribute trends. The effectiveness and efficiency of the DTSC algorithm are validated by experiments on simulated datasets and real applications. In the applications of simulated datasets, the results indicate that the proposed DTSC algorithm effectively detects time series clusters with arbitrary shapes and similar attributes and densities while considering noises. In the real applications, both time series raining dataset and time series surface deformation dataset have been utilized, and several interesting patterns that cannot be effectively detected by other classical time series clustering algorithms have been found.