An integrated approach of neural network and decision tree to classification

This paper present a new integrated approach of neural network and decision tree to classification. Neural network (NN) are frequently applied to classification with various objectives. But its knowledge is in the networks, which can´t be apprehensible. Decision tree (DT) is one of the most popular approaches for classification, it can extract comprehensible rule based on the training dataset, but the tree maybe too big when the features too more and the databases are too big. Therefore, in this paper, the NN is used to reduce the irrelevance feature set and filter the noise data in the training dataset. Then, condensing the training set by clustering them. The DT extract rule set from the worked training dataset. It can enhance the precision of classification, generalization performance and reduce the number of rule. The experiment demonstrates the effectiveness of the mentioned algorithm.