Evaluating the Efficiency of the Radial Basis Function (RBF) Neural Network Model in Forecasting Economic Time Series: An Applied Study

Abstract

This study aimed to evaluate the effectiveness of the Radial Basis Function (RBF) neural network model in forecasting a time series of economic or statistical variables during the period from July 2009 to June 2011, relying on six independent variables (X1–X6) to estimate the dependent variable (RBF). The results demonstrated that the model possesses high predictive accuracy, as indicated by the Pearson correlation coefficient between actual and predicted values for key variables such as X1, which reached approximately 0.957, with a high statistical significance (p < 0.01). The predicted values from the model reflected an upward trend in the RBF variable, aligning with the temporal patterns in the data. Additionally, variables X2 and X3 emerged as relatively important predictors. The model showed strong performance levels according to statistical quality metrics such as AIC, SIC, and HQC, alongside a high coefficient of determination (R²). Despite the overall efficiency of the model, some discrepancies in prediction accuracy were observed, particularly with variables X3 and X5, which showed relatively high values in MAPE and TS, indicating potential for improving the model’s performance in these cases.

Based on the findings, the study recommends the following:

• Enhancing the use of the RBF model in statistical and economic forecasting applications.
• Reviewing the performance of certain variables to improve predictions.
• Regular application of model evaluation criteria.
• Improving data quality and training technical teams on the use of such models.

This study confirms the effectiveness of the RBF model as a powerful tool in time series analysis and accurate forecasting, making it a suitable option in complex and dynamic environments.