Bayesian Model-Assisted PRIM Algorithm
Bayesian Model-Assisted PRIM Algorithm
复制标题
贝叶斯模型辅助 PRIM 算法
DOI：
--
复制
发表时间：
2002
期刊：
影响因子：
--
通讯作者：
H. Chipman
中科院分区：
文献类型：
--
作者：
Longyang Wu;H. Chipman
患者规则归纳法（PRIM）是一种统计学习方法，旨在定位特征空间中响应变量具有高值的区域。在本文中，我们提出了一个贝叶斯模型辅助PRIM算法。该算法能够根据贝叶斯因子值和边缘后验概率自动覆盖特征空间中有希望的区域。该方法的模型辅助部分开始与定量响应变量的灵活的均值-方差移动模型的规格。这使得预测成为可能，并且有可能通过贝叶斯模型平均来实现更高的预测能力。最后给出了一个仿真例子来说明该算法的有效性。
The patient rule-induction method (PRIM) is a statistical learning method that seeks to locate regions in the feature space where the response variable has a high value. In this paper we present a Bayesian model-assisted PRIM algorithm. This algorithm can automatically cover promising regions in the feature space based on Bayes factor values and marginal posterior probabilities. The model-assisted part of the method begins with the specification of a flexible mean-variance shift model for quantitative response variables. This enables predictions, and has the potential to achieve even higher prediction power through the Bayesian model averaging. A simulated example is provided to illustrate the effectiveness of this new algorithm.