Abstract: Through researches and analyses on Multi-Relational Decision Tree Learning(MRDTL) algorithm, MRDTL-2, for the problems of low running efficiency and poor effect in dealing with missing attribute values of it, this paper introduces an Improved Multi-Relational Data Mining (IMRDTL) algorithm. IMRDTL algorithm makes use of the tuple ID propagation for speeding up the efficiency of MRDTL-2 algorithm, meanwhile, generalized naive Bayes classifier is used to fill in missing attribute values, so the algorithm can provide a better accuracy.

Key words: Multi-Relational Data Mining(MRDM), decision tree, tuple ID propagation, generalized naive Bayes

摘要： 通过对多关系决策树学习算法MRDTL-2进行研究与分析，针对其运行效率较低和不能有效处理丢失属性值的问题，提出一种改进的多关系数据挖掘(IMRDTL)算法。在IMRDTL算法中，利用元组ID传播技术来进一步提高MRDTL-2算法的运行效率，同时使用广义朴素贝叶斯分类器来填补丢失的属性值，以进一步提高算法的准确率。

关键词: 多关系数据挖掘, 决策树, 元组ID传播, 广义朴素贝叶斯

CLC Number: 

• TP301.6