一种基于神经网络和决策树相结合的数据分类新方法

提出了一种将神经网络和决策树相结合的数据分类新方法。该方法首先依据属性重要性将属性进行排序,然后通过RBF神经网络进行属性裁减,最后生成决策树,并抽取出规则。与传统的决策树分类方法相比,此方法可依据属性重要性直接生成最小决策树,避免了树的裁减过程,大大加快决策树的生成效率,并进一步提高了规则的预测精度。该方法适用于大规模及高维属性的数据分类问题。

A New Data Classifying Method Based on Combination of Neural Network and Decision Tree

This paper presents a new data classifying method based on combination of neural network and decision tree. The method firstly ranks attributes based on the importance of the attributes, and then prunes the attributes using RBF neural network, and finally builds a decision tree and extracts rules. Compared with the traditional data classifying methods using decision tree, the present method can gain the minimal decision tree directly without pruning, which largely raises the efficiency of building decision tree and improves the prediction precision of rules produced. The method is suitable for large scale and high dimension data classifying problem.