A decision cluster classification model with model selection strategy

The Decision Cluster Classification (DCC) model is a clustering-based classification method. The aim is to solve the hard problem in classification with the aid of clustering techniques. In DCC model, model selection is an important step, which optimizes DCC model in classification accuracy and time consuming. This paper focuses on the model selection problem and proposes a scheme. Firstly, a theory is proposed to estimate the upper error bound for a DCC model. Then, we give a strategy to select the best DCC model by reducing error bound. Empirical results show that our theory is reasonable and our model selection method can enhance DCC model. Moreover, the optimized DCC model can outperform famous classifiers (e.g. SVM, C4.5) on some data sets.