基于Stacking集成学习算法的个人信用评估模型.pdfVIP

Statistics and Application 统计学与应用, 2017, 6(4), 411-417 Published Online October 2017 in Hans. /journal/sa /10.12677/sa.2017.64047 Personal Credit Assessment Model Based on Stacking Ensemble Learning Algorithm Runze Peng School of Mathematics and Systems Science, Beihang University, Beijing th th th Received: Oct. 4 , 2017; accepted: Oct. 19 , 2017; published: Oct. 26 , 2017 Abstract The prediction accuracy of traditional machine learning methods often depends on the specific problems. Ensemble learning achieves significant improvement in classification performance by combining several of base classifiers. This paper briefly introduces the basic idea of ensemble learning, discusses advantages of Stacking to the traditional classical ensemble algorithms. Based on the Stacking framework，we build two-layer classification model to evaluate the personal credit using the UCI datasets. The results of the empirical analysis show that, compared with the single machine learning method and simple average ensemble, Stacking with two-layer classifier has a better prediction effect. Keywords Ensemble Learning, Stacking, Credit Assessment 基于Stacking集成学习算法的个人信用 评估模型 彭润泽 北京航空航天大学数学与系统科学学院，北京 收稿日期：2017年10月4 日；录用日期：2017年10月19 日；发布日期：2017年10月26 日 摘 要 传统机器学习算法的预测精度往往依赖于具体的问题，集成学习通过综合若干基分类器的预测结果，实 文章引用: 彭润泽. 基于Stacking 集成学习算法的个人信用评估模型[J]. 统计学与应用, 2017, 6(4): 411-417. DOI: 10.12677/sa.2017.64047 彭润泽 现了分类效果的显著提升。对集成学习的思想进行了简单地介绍，阐述了Stacking集成相对于传统经典 集成算法的优势。并基于Stacking集成框架，利用UCI的信用评估数据集，构建两层分类器学习模型对个 人信用进行评估。实证分析的结果表明，相对于单一的机器学习方法，以及对这些单一机器学习方法的 结果进行简单的平均集成，两层分类器的Stacking集成学习有着更好的预测效果。 关键词 集成学习，Stacking，信用评估