中文摘要：

以我国证券市场涉及到利润和资产造假的30家上市公司作为研究样本，利用GMDH模型对我国上市公司虚假财务报告进行识别。研究得到了可以显示财务造假征兆的5个关键异动指标，并且，通过其中2个指标可以很好识别那些以虚减利润和收入的造假方式为管理层输送利益的上市公司。实证研究表明，GMDH模型具有较高的识别正确率和较强的预测推广能力。

英文摘要：

According to the contrast of the financial index of 30 listed companies in China having appearance of skullduggery during 1996 -2005, by advanced data mining technology based on the method of GMDH （Group Method of Data Handling）, this paper establishes GMDH models in order to identify skullduggery of financial reports. The results prove that these GMDH models are valid to identify skull- duggery of financial report.