神经网络与领域知识结合的纳税评估预警模型

Risk Warning Model of Tax Payment Evaluation Based on Neural Network and Domain Knowledge

摘要: 纳税评估是一项重要而复杂的工作。针对目前尚无十分有效的纳税评估预警模型的情况，提出了将神经网络和领域知识结合建立纳税评估预警模型的方法，利用基于神经网络的方法选出有涉税疑点的企业，在领域知识的指导下结合统计分析方法，解决了预警模型无疑点指向性的问题。通过建立行业的纳税评估预警模型，并进行验证分析，表明该方法是可行的。

Abstract: A risk warning model of tax payment evaluation based on neural network and domain knowledge is presented in this paper. The enterprises that have the most possibility of tax dodging can be found by applying neural network methods. The doubtful points in tax payment of these enterprises are evaluated by statistical analysis methods. A case study shows the feasibility of the model.