摘要
为了提高企业信用风险评估准确率,提出了基于PSO-BP集成的企业信用风险评估模型。使用Bagging抽样技术获得足够多不同的训练数据集,用不同的训练集子集训练得到不同的PSO-BP组合成员分类器,最后通过多数投票准则整合不同组合成员分类器的分类结果。分别在包含了国内外公司的详细数据的数据集上证明了模型的有效性。
This paper proposed a PSO-BP ensemble approach for enterprise credit risk assessment. In this proposed model, it first used a Bagging sampling approach to generate different training sets for guaranteeing enough training data. In terms of dif- ferent training datasets, it trained muhiple individual PSO-BP classifiers. Finally it aggregated the ensemble members in terms of some criteria and output their generated results based upon majority vote rules. Experiment on the data set which contains the detailed information of domestic and foreign corporations proves the validity of the ensemble model.
出处
《计算机应用研究》
CSCD
北大核心
2014年第9期2705-2710,共6页
Application Research of Computers
基金
上海市科学技术委员会科研计划资助项目(13dz1508400)
上海财经大学研究生创新基金资助项目(CXJJ-2012-322)
国家自然科学基金资助项目(71301095)