CPI Inflation Rate: A Comparative Analysis of Traditional Basic Structural Model (BSM) and Artificial Neural Networks (ANN) Model for Accurate Predictions in South Africa
Preprint
- 21 January 2026
- preprint
- Published by Springer Nature in Research Square
Abstract
Precise inflation prediction is crucial for the formulation of monetary policies and economic planning, especially in the case of developing countries like South Africa. This research investigates the predictive capabilities of conventional and machine learning models for forecasting the South African Consumer Price Index (CPI) inflation rate. The monthly values of the CPI inflation rate from January 2008 to October 2025, collected from Statistics South Africa, were considered. The models used were the Basic Structural Model (BSM), the Artificial Neural Network (ANN), and the combined hybrid BSM and ANN approach. The BSM model examines structural variables like trend and seasonality, and the model ANNs focus on nonlinearities. The model performance was assessed using the criteria of RMSE, MAE, and MASE performance measures, in the in-sample forecasting. The empirical analysis shows that the performance of the model ANN outperformed that of the BSM and hybrid BSM–ANN models considerably when compared using the in-sample forecasting, and its results were well-behaved without any features of autocorrelation and heteroscedasticity. The results showed that the combined hybrid BSM–ANN model was superior to the BSM and ANN models, although it may not necessarily outperform the other two, as the out-of-sample forecasts were more stable and reasonable.Keywords
All Related Versions
This publication has 19 references indexed in Scilit:
- Inflation expectations and monetary policy: what have we learned and to what end?The Econometrics Journal, 2026
- Forecasting CPI inflation under economic policy and geopolitical uncertaintiesInternational Journal of Forecasting, 2024
- The consumer price index prediction using machine learning approaches: Evidence from the United StatesHeliyon, 2023
- An investigation of artificial neural network structure and its effects on the estimation of the low‐cycle fatigue parameters of various steelsFatigue & Fracture of Engineering Materials & Structures, 2023
- Explainable inflation forecasts by machine learning modelsExpert Systems with Applications, 2022
- Forecasting CPI inflation components with Hierarchical Recurrent Neural NetworksInternational Journal of Forecasting, 2022
- Time Series Analysis by State Space Methods2012
- Neural network models for inflation forecasting: an appraisalApplied Economics, 2011
- Modelling and Forecasting Tourist Flows to Barbados Using Structural Time Series ModelsTourism and Hospitality Research, 2010
- Time series forecasting using a hybrid ARIMA and neural network modelNeurocomputing, 2003