一种并行决策树学习方法研究

提出一种并行的决策树学习方法。该方法首先将数据库分为若干个子部分,针对每个子部分分别进行决策树学习,优选分裂属性,再对各个决策树学习的结果综合,生成最终的树。在树上剪枝以降低分类错误率。通过在变压器绝缘故障诊断中的应用表明该方法有很强的学习能力和诊断速度,是一种有效的决策树学习方法。

A Parallel Decision Tree Learning Approach

A new parallel learning method of the decision tree is proposed in this paper.The method firstly divides the overall knowledge base into several subsections,then based on every such subsection a decision tree is assigned to learn to select the splitting attribute,respectively,the results learned by each decision tree are incorporated to generate the final tree.The tree would be pruned to reduce classification error and over-fitting.Practical application in transformer insulation fault diagnosis shows that the proposed method possesses very strong learning ability and diagnosis speed,and is an effectively learning method of decision tree.