Advanced AI models for future forecasting of budget expenditures via machine learning and deep learning

This study builds on T?rkiye?s long-standing challenges in managing public spending amid political and economic uncertainty. Budget planning plays a vital role in ensuring fiscal sustainability and economic resilience. Therefore, this study proposes a new forecasting framework that utilizes the latest artificial intelligence models. This paper aims to provide data-driven decision support to policymakers by improving the accuracy and robustness of expenditure forecasts under complex temporal dynamics. Accordingly, the use of machine learning and deep learning methods to forecast budget expenditures in T?rkiye was proposed and analyzed. Comprehensive datasets extending from January 2008 to May 2024 was considered. Datasets also include various extraordinary periods such as the global financial crisis, the European debt crisis, various major political events in T?rkiye, and the COVID-19 pandemic. Model performance was evaluated using the Time Fusion Transform, which has received praised for its superior performance even in complex and volatile time series. The model results show that the MAPE is 0.0658%, MAE is 0.0050%, RMSE is 0.0111%, and R? is 0.993%. The Random Search learning algorithm was implemented to determine the optimal hyperparameters that enable the model to work effectively on the data. According to the findings of the study, the model can perform well despite economic and political changes. Biodiversity in the use of all model types shows that machine learning and deep learning models also offer valuable insights into the budget forecasting process.