基于组合分类的消费者信用评估

发布时间：2018-05-13 12:46

本文选题：有指导聚类 + 组合分类　；参考：《管理工程学报》2015年01期

【摘要】：为了降低单个分类模型的不稳定性,提高其应用于消费者信用评估的准确性,提出一种基于组合分类的消费者信用评估方法。首先通过有监督聚类将各个类别的数据样本划分为若干子集合,使得各个子集合中数据样本均来自同一类别,再对不同类别子集合之间进行两两组合得到训练样本子集合,然后在各个训练样本子集合中分别建立不同的分类模型。在分类模型结果综合阶段,以各个分类模型在待分类样本的近邻训练样本中的分类性能作为权重,对各个分类模型的结果进行加权投票以产生待分类样本的最终分类结果。实证研究以决策树作为基本分类器,通过在实际的消费者信用数据集上的比较分析,说明所提出方法相对于其它组合分类方法具有更高的分类准确性,可以有效应用于消费者信用评估。
[Abstract]:In order to reduce the instability of a single classification model and improve the accuracy of its application to consumer credit evaluation, a method of consumer credit evaluation based on combinatorial classification is proposed. First, the data samples of each category are divided into several subsets by supervised clustering, so that the data samples in each subset are all from the same category, and then the training sample subsets are obtained by pairwise combination of different subsets of categories. Then, different classification models are established in each subset of training samples. In the stage of classification model result synthesis, the classification performance of each classification model in the nearest neighbor training sample of the sample to be classified is taken as the weight, and the results of each classification model are weighted to vote to produce the final classification result of the sample to be classified. Taking decision tree as the basic classifier, the empirical study shows that the proposed method has higher classification accuracy than other combined classification methods through comparative analysis on the actual consumer credit data set. It can be applied to consumer credit evaluation effectively.
【作者单位】：重庆大学经济与工商管理学院;重庆大学现代物流重庆市重点实验室;
【基金】：国家自然科学基金项目资助项目(71001112) 中央高校基本科研业务费资助项目(CDJSK 11066)
【分类号】：F832.4;F224

【相似文献】