一种新的增量决策树算法

对于数据增加迅速的客户行为分析、Web日志分析、网络入侵检测等在线分类系统来说,如何快速适应新增样本是确保其分类正确和可持续运行的关键。该文提出了一种新的适应数据增量的决策树算法,该算法同贝叶斯方法相结合,在原有决策树的基础上利用新增样本迅速训练出新的决策树。实验结果表明,提出的算法可以较好的解决该问题,与重新构造决策树相比,它的时间开销更少,且具有更高的分类准确率,更适用于在线分类系统。

A New Algorithm for Incremental Induction of Decision Tree

Incremental learning is an effective method for learning the classification knowledge from massive data,especially in the situation of high cost in getting labeled training examples.According to the characteristic and essence of incremental learning,this paper gives a new incremental decision tree classification algorithm which combined Bayesian method and decision tree.The design and test of the algorithm was introduced too.The experimental result shows that this classification model retains the good interpretability of decision tree and has the ability of incremental learning.