Tax Risk Prediction of Real Estate Based on Convolutional Neural Network

Abstract

Risk management is an important link in tax administration. From China’s taxation practice, risk identification has become the weakness of tax management. With the complexity of massive data and the secrecy of modern transactions, traditional tax risk identification can no longer adapt to the development of the times. In the past, most risk researches focused on the basic machine learning stage. There are gaps in the application of deep learning in tax risk management. Based on the tax risk management indicators, this paper took the real estate industry as an example. We used convolutional neural network (CNN) to construct a tax risk prediction model. The experiment shows that a tax risk prediction model based on CNN has higher accuracy in tax risk identification and has a stronger ability to process tax data. The model has a certain reference value for tax authorities to reduce tax risk and tax loss.