An Unsupervised Grey Incidence Clustering Method Based on Multi-Dimensional Dynamic Time Warping Distance

Grey incidence clustering is an effective tool for uncertainty information system processing. Existing grey incidence clustering methods can only deal with one-dimensional and complete sequences. This results in a lot of extra works to be done, such as alignment sequences and assignment cluster threshold artificially. It will greatly affect the clustering performance. To solve these problems, an unsupervised grey incidence clustering method based on multi-dimensional dynamic time warping distance is proposed. It measures the incidence degree between multi-dimensional sequences by calculating the minimum distance of a dynamic warping path. On the foundation, an improved k-means clustering method is constructed to realize grey incidence clustering by the strategy of gradually increasing the number of categories. The experiment result shows that the proposed method has a perfect performance on clustering especially on dealing with multi-dimensional or incomplete sequences.