C-fuzzy variable-branch decision tree with storage and classification error rate constraints

摘要

The C-fuzzy decision tree (CFDT), which is based on thenfuzzy C-means algorithm, has recently been proposed. The CFDT isngrown by selecting the nodes to be split according to its classifica-ntion error rate. However, the CFDT design does not consider thenclassification time taken to classify the input vector. Thus, the CFDTncan be improved. We propose a new C-fuzzy variable-branch deci-nsion tree (CFVBDT) with storage and classification error rate con-nstraints. The design of the CFVBDT consists of two phases—ngrowing and pruning. The CFVBDT is grown by selecting the nodesnto be split according to the classification error rate and the classifi-ncation time in the decision tree. Additionally, the pruning methodnselects the nodes to prune based on the storage requirement andnthe classification time of the CFVBDT. Furthermore, the number ofnbranches of each internal node is variable in the CFVBDT. Experi-nmental results indicate that the proposed CFVBDT outperforms thenCFDT and other methods.