基于支持向量机的高新技术企业与传统企业的类型辨识模型研究

Type Identification Model of High-tech Companies and Traditional Companies

基于支持向量机神经网络理论,首创性地建立了一个由业绩产出财务指标辨识高新技术企业与传统企业类型的支持向量机模型。模型以企业的业绩产出财务指标数据为基础,以径向基函数作为核函数,使用网格寻优方法调节模型参数,得到优化后的模回去型,并使用测试集数据验证了模型。对结果进行二元分类决策分析,结果表明:该模型的准确率和决策率等主要评价指标都达到了85%以上,具有较高的辨识能力和可信度,为高新技术企业和传统企业的类型辨识提供了一种可靠的、简单方便的方法,可以直接量化地判别企业是否属于高新技术企业。

This paper presents a novel identification model for the identification of high technology companies and traditional ones from financial performance indexes for the first time, based on the support vector machine （SVM） neural network （NN）. The model is on the basis of the data of companies＇ indexes, employs radial basis function （RFB） as the kernel function. The kernel parameters are selected and adjusted by grid search method. The optimized model is verified by the test data. The results are discussed by binary classification decision analysis. It indicates that the accuracy, precision, recall and other main evaluation indexes of the model are achieved 85% above, which means high reliability. The model provides a reliable, simple and convenient approach for the type identification of high technology companies quantitatively.