Data Mining Using Artificial Neural Networks and Convolutional Neural Networks for Predicting Tax Revenues: An Applied Study at the General Authority for Taxes

Abstract

This research aims to employ Artificial Neural Networks (ANN) and Convolutional Neural Networks (CNN) techniques in forecasting tax revenues, which represent one of the most important financial resources that the state relies on to finance its public expenditures and support its economic policies. The research adopts both the descriptive-analytical approach and the quantitative approach in analyzing time series data of tax revenues issued by the General Commission for Taxes in Baghdad for the period (2003–2025), with future forecasts conducted for the years (2026–2030). Forecasting models were implemented using (MatLap-2023), and their performance was evaluated based on a set of statistical accuracy measures, including Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), in addition to the correlation coefficient. The results of the analysis revealed that the Artificial Neural Network (ANN) model outperformed the Convolutional Neural Network (CNN) model in terms of forecasting accuracy, achieving lower error values along with a higher correlation coefficient between actual and predicted values. Furthermore, the forecasting results for the future period indicated a relative stability in the trend of tax revenues, with some variation in the predicted values between the two models. The study concludes that the application of data mining techniques, particularly Artificial Neural Networks, is an effective tool for improving the accuracy of tax revenue forecasting, thereby supporting financial decision-makers and enhancing the efficiency of financial planning and public budget preparation.