摘要
应用自组织映射神经网络(SOM)和Hopfield神经网络模型对上市公司并购目标公司进行了实证研究。SOM网络的聚类分析表明目标公司可分为6个类别,各类别之间差异较大,目标公司明显区别于非目标企业,在总体上具有盈利能力低、经营能力差、偿债能力较强的特点。Hopfield网络模型的预测结果显示,目标企业的平均预测准确率为80.69%,非目标企业的预测准确率为61.33%,由于并购交易发生受多种因素影响,财务指标与其它因素相结合方能提高模型预测的效果。
In this paper, the authars apply SOM and Hopfield neural network to cluster and predict the target of mergers and acquisitions （M＆A）. Financial characteristics of six sorts of targets are shown with low profitability, bad operation and good solvency very evidently by clustering of SOM. After calculating the means of variables of every sort, we build Hopfield network to predict the sort of targets and non-targets according to the means. Demonstration indicates Hopfield network can be used as prediction although accuracy of target selection is 80. 69%, and non-target is 61.33% on the average. Financial index should be combined with other index to improve prediction of the model.
出处
《统计与信息论坛》
CSSCI
2010年第6期58-62,共5页
Journal of Statistics and Information
基金
国家自然科学基金项目<局内不确定性政策理论及其在交通运输管理中的应用研究>(70671004)
苏州市社会科学基金项目<县域经济保增长与优化升级问题研究>(09-C-33)
关键词
并购
目标公司
财务特征
SOM网络
HOPFIELD网络
Mergers and Acquisitions target company financial characteristic Self-organized Feature Mapping Neural Network
Hopfield Neural Network
