Dissimilarity based on direction information and its application

Similarity (dissimilarity) is of critical importance for data analysis, especially for clustering problem. The classical dissimilarity is related to distance (norm of the difference between each pair of data points), but it ignores the direction information from one data point to another. In this paper, we proposed a new dissimilarity based on direction consistence, which considers not only the distance information but also the direction information. It has some advantages and can be used in clustering to give a good performance.