Internal Cluster Quality Indexes for Classification of Symbolic Data
Andrzej Dudek · University of Lodz Repository (University of Łódź) · 2009
This paper describes main classification methods used for symbolic data (e.g. data in form of: single quantitative value, categorical value, interval, multivalued variable, multivaliued variable with weights) presents difficulties of measuring clustering quality for symbolic data (such as lack of "traditional" data matrix), presents which of known indexes like Silhouette index, Ball index, Hartingan index, Baker and Hubert index, Huberta and Levine index, Ratkovski index, Ball index, Hartigan index, Krzanowski and Lai index, Scott index, Marriot index, Rubin index, Friedman index may be used for validation of such type of data and what indexes are specific only for symbolic data. Simulation results arc used to propose most adequate indexes for each classification algorithm.