Clustering Mixed Data via Diffusion Maps

Data clustering is a common technique for statistical data analysis, which is used in many fields, including machine learning, data mining, customer segmentation, trend analysis, pattern recognition and image analysis. Although many clustering algorithms have been proposed most of them deal with clustering of numerical data. Finding the similarity between numeric objects usually relies on a common distance measure such as the Euclidean distance. However, the problem of clustering categorical (nominal) data is more difficult and challenging since categorical data have nominal attributes. As a result, finding the similarities between nominal objects using common distance measures, which are used for processing numeric data, is not applicable here. Moreover, real applications data have to deal with mixed types of attributes such as numeric and nominal data that reside together. In this paper, we propose a technique that solves this problem. We suggest to transform the input data (categorical and numerical) into categorical values. This is achieved by an automatic non-linear transformations, which identify geometric patterns in these datasets, that find the connections among