摘要
通过线性组合构造混合核函数,建立一种基于混合核学习的支持向量机财务欺诈检测模型。利用蜂群算法对混合核函数参数进行寻优,获取最佳参数,并对给定的训练样本进行学习,得出最佳输入输出关系,从而对财务数据进行识别检测。实例测试结果表明,该模型与单核的支持向量机模型相比,识别精度和鲁棒性都有所提高。
A mixture kernel learning method is proposed to construct a model of financial fraud detection based on mixed kernel learning and support vector machine. The optimal parameters of this machine are obtained by the improved bee colony algorithm. The optimal input and output relations of the given training samples are obtained so as to carry on the recognition examination to the financial data. Experimental results show that the proposed model can improve the recognition accuracy and robustness compared with the single-kernel SVM model. The effectiveness of the algorithm is verified by experiments. Compared with the traditional support vector machine model, the recognition accuracy and robustness are improved.
出处
《西安邮电大学学报》
2017年第2期81-83,88,共4页
Journal of Xi’an University of Posts and Telecommunications
基金
国家自然科学基金重点资助项目(61136002)
陕西省自然科学基金资助项目(2014JM8331
2014JQ5183
2014JM8307)
陕西省教育厅科学研究计划资助项(2015JK1654)
关键词
混合核函数
支持向量机
蜂群算法
财务欺诈
mixture kernel function, support vector machine, bee colony algorithm, financial fraud
