Entropy-Based Mixed Data Transform Model

摘要

In the era of big data, in-depth data mining is inevitable and urgent. Cluster analysis is one of the major data mining methods. Measuring similarity or distance between two objects is a key step for several data mining and knowledge discovery tasks. For abound unstructured free data, conversion between numeric data and categorical data matters most, while the notion of similarity for numeric data is relatively well-studied and for categorical data not satisfying. Learning from current clustering algorithm for categorical data and mixed data, several methods and corresponding features are explored and summarized. Results on a variety of data sets show that while no one measure dominates others for all types of problems, but some measures are able to be integrated into clustering process. Proposed method has the potential capability to deal with numeric and categorical features (mixed features) of dataset.