改进的多关系决策树算法  

Improved multi-relational decision tree algorithm

作　　者：宋广玲[1] 郝忠孝[1] 吴海燕[1] 

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《计算机应用研究》2009年第12期4502-4505,4512,共5页Application Research of Computers

基　　金：黑龙江省自然科学基金资助项目(F200702)

摘　　要：高效性和可扩展性是多关系数据挖掘中最重要的问题,而提高算法效率的主要瓶颈在于假设空间,且用户对分类的指导会在很大程度上帮助系统完成分类任务,减少系统独自摸索的时间。针对以上问题提出了改进的多关系决策树算法,即将虚拟连接元组传播技术和提出的背景属性传递技术应用到多关系决策树算法中。对改进的多关系决策树算法进行了理论证明,并且对多关系决策树算法和改进的多关系决策树算法进行比较实验。通过实验可以得出,当改进的多关系决策树在搜索数据项达到背景属性传递阈值时,改进的多关系决策树算法的效率相对很高且受属性个数增加(或记录数增加)影响较小。因此提出的算法优于现有的同类算法,实现了预期的研究目标。Efficiency and sealability of data mining has been an important problem in muli-relational data mining. For muhirelational data learning algorithm, the main bottleneck of improving algorithm efficiency is the assumption space. And the u-ser＇s guide for the classification will help the system to complete classification task in a large extent, reduce the time of searching by system alone. In view of the above problems, this paper proposed the improved multi-relational decision tree algorithm： tuple ID propogation technologies of virtual and proposed a technology of the delivery of background attributes connecting applications to multi-relational decision tree algorithm. Gave the theoretical proof and proving of comparing experiment between algorithm of multi-relational decision tree and improved algorithm of multi-relational decision tree. Through the above experimental, it can be drawn that, when the improved multi-relational decision tree algorithm meets the threshold of the background of property transfer in the search data item, the improved algorithm based on multi-relational decision tree has relatively higher operating efficiency and is less affected by increasing the number of property, so the proposed algorithm is better than existing similar algorithms, it achieves research objectives.

关 键 词：多关系数据挖掘 多关系决策树 元组标志传播 背景属性 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP301.6[自动化与计算机技术—计算机科学与技术]