An improved algorithm for DDAGSVM

Decision directed acyclic graph support vector machine (DDAGSVM) is an effective approach to solve multi-class problem, but it has to solve the problem of how to choose the structure of the graph and minimizing the classification error that might be accumulated at the final classification process. In order to improve the generalization ability of DDAGSVM, and minimizing the classification error that might be accumulated at the final classification process, the efficient method is studied in this paper. Based on the idea that the most separable classes should be separated firstly during the formation of DDAG, and the effective class separability measure should take the distribution of the classes into consideration, a separability measure is defined based on the distribution of the training samples in the kernel space, and by introducing the defined between-class separability measure into the formation of DDAG, an improved DDAGSVM algorithm is given. Classification results for the data sets prove the effectiveness of the improved DDAGSVM