【摘要】 分析被审计单位数据从而检测出欺诈记录是当前审计工作的一个重要课题,传统的数据挖掘方法在处理该问题时存在很大的局限性。论文提出了一种基于免疫网络的分类算法,基于训练数据构建自我和非我网络来提取正常模式和欺诈模式。算法根据新数据同自我非我网络的匹配情况来定量地计算欺诈分来实现分类。算法引入了免疫学习、免疫克隆、免疫记忆机制,并引入免疫变异机制提高对未知模式的识别能力。论文针对标准数据和审计数据完成了相应的验证实验。结果表明该算法具有较好的分类能力和欺诈检测能力。更多还原

【Abstract】 Analysizing electronic data of audited organ to detect fraud is a important problem of current audit-work.Conventional technique of data mining have many shortcomings for this question.So,this paper designes an classification algorithm which is based on immune network theory:The algorithm builds self-network and negative-network to respectively extract natural pattern and fraud pattern,then compares the match degree of new data and two networks to compute its "fraud score".The algorithm inducts immune learning,clone,memory and mutation mechanism to improve its ability for unknown patterns.This paper accomplishes corresponding experiments for both standard data and audit data.The experiment results are given and show that this algorithm had a good ability to classification and fraud detection.更多还原