Forecasting the Inflation for Budget Forecasters: An Analysis of ANN Model Performance in Türkiye

The reliability of budget revenue and expenditure forecasts depends on the accuracy of inflation forecasts. Without realistic inflation forecasts, it is not possible to produce sound budget forecasts. This study aims to guide budget forecasters in Türkiye by providing accurate inflation forecasts. The analysis utilizes data from the 2005–2023 period. The basket exchange rate (USD and Euro), unemployment, imports, exports, budget revenues and expenditures, interest rates, industrial production index, money supply, general price index, and minimum wage are forecasted using Holt-Winters, ARIMA, SARIMA, Prophet, LSTM, and Hybrid models. These forecasts are then used as inputs in ANN, SVR, RF, and GBM models to forecast monthly inflation. The results indicate that the forecasts generated with ANN are significantly more realistic than those presented in Türkiye’s budget law and the Medium-Term Program. The study demonstrates that ANN can be an effective tool for budget forecasters in accurately forecasting inflation and, consequently, improving budget forecasts. The findings are further evaluated through a comparative analysis with forecasts from institutions such as the IMF, OECD, Central Bank, and the European Union. To support future academic research, inflation forecasts for 2025, along with forecasts for independent variables, are also included in the study.