摘要
在数据挖掘领域中对于测试例代价和概率未知的情况下,进行代价估计和概率估计是必须的.研究直接Cost-Sens itive决策方法的思想和决策过程,并运用决策树和贝叶斯学习方法进行代价估计和概率估计.当运用以上方法之后,能得到被校准的代价估计.与M etaCost方法相比较,直接Cost-Sens itive决策方法是更加有效的决策方法.
When both costs and probabilities are unknown in test cases in the field of data mining,it is necessary to estimate cost and probability. A thought and decesion-making process of a direct Cost-Sensitive decision-making method are proposed. Decision tree and Bayesian learning methods are used in cost estimating and probability estimating. Based on the above methods ,the cost can be estimated by calibration. It shows that the direct Cost-Sensitive decision-making method performs better than Metacost method.
出处
《内蒙古大学学报(自然科学版)》
CAS
CSCD
北大核心
2009年第4期465-468,共4页
Journal of Inner Mongolia University:Natural Science Edition
关键词
代价
概率
分类算法
决策
cost
probability
classification algorithm
decision