Search

Common goal of descriptive data mining techniques is presenting new information in concise, easily interpretable and understandable ways. Hierarchical clustering technique for example enables simple visualization of distances between analyzed objects or attributes. However, common distance measures used by existing data mining tools are usually not well suited for analyzing transactional data using this particular technique. Including new types of measures specifically aimed at transactional data can make hierarchical clustering a much more feasible choice for transactional data analysis. This paper presents and analyzes convenient measure types, providing methods of transforming them to represent distances between transaction elements more appropriately. Developed measures are implemented, verified and compared in hierarchical clustering analysis on both artificial data as well as referent transactional datasets.

Common goal of descriptive data mining techniques is presenting new information in concise, easily interpretable and understandable ways. Hierarchical clustering technique for example enables simple visualization of distances between analyzed objects or attributes. However, common distance measures used by existing data mining tools are usually not well suited for analyzing transactional data using this particular technique. Including new types of measures specifically aimed at transactional data can make hierarchical clustering a much more feasible choice for transactional data analysis. This paper presents and analyzes convenient measure types, providing methods of transforming them to represent distances between transaction elements more appropriately. Developed measures are implemented, verified and compared in hierarchical clustering analysis on both artificial data as well as referent transactional datasets.