基于递归神经网络的纳税评估预警模型

Tax Assessment Warning Model Based on Recurrent Neural Network

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,提出利用递归神经网络(RNN)建立纳税评估预警模型的方法,利用RNN的方法选出有涉税疑点的企业,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,表明该方法可行。

Tax assessment is an important and complicated task. Under the situation that there isn’t a very effective warning model for tax assessment at present, this paper points out a way of using recurrent neural network (RNN) to establish warning model of tax assessment, and select the enterprises which have some tax suspects. This can solve the problem that warning model has no directive property of suspects. Through establishing warning model of tax assessment in industries, and verifying the analysis, this paper shows that this method is feasible.