Multi-Level Support Vector Machine

Many type of classifiers have been presented in machine learning and large margin classifier is one of them. Support Vector Machine (SVM) is the most famous of large margin classifiers. SVM is a very useful classifier, but has some limitations. Only patterns near the decision boundary are used as support vectors and decision making in SVM is done locally. In this paper we propose a method that also uses information in other patterns for classification which is called "Multi-Level Support Vector Machine" (MLSVM). We compare our method with the original SVM and an artificial neural network using some UCI datasets and the final results shows that our method is better or equal to other methods.