Abstract: Accurate credit risk assessment is very important for risk management of financial institutions. In order to further improve the accuracy of credit risk assessment model, a hybrid ensemble approach, RSA-SVM, is proposed. This approach is a SVM-based ensemble learning method, combined with two popular ensemble strategies, random subspace model and AdaBoost, in order to improve the diversity of base learners, and get a more accurate ensemble learning model. This model is applied on two public credit data sets, and the experimental results show that the RSA-SVM is a more effective model for credit risk assessment.

Key words: credit risk assessment;　Support Vector Machine（SVM）, ensemble learning, AdaBoost, random subspace model

摘要： 准确的信用风险评估可以降低金融机构的风险。为了进一步提高信用风险评估模型的预测准确率，将基于SVM的集成学习模型应用到信用风险评估问题中，提出了一种混合集成策略，称作RSA。RSA是随机子集模型和AdaBoost两种流行策略的合成，能提高组合成员分类器的多样性，从而提高集成学习模型的预测准确率。模型在两组公开信用数据集上进行了应用，实验结果表明基于RSA的SVM的集成学习模型可以作为信用风险评估的有效模型。

关键词: 信用风险评估, 支持向量机（SVM）, 集成学习, AdaBoost, 随机子集模型