Inflation prediction: An hybrid time series approach
Authors
Ameh, Tonia
Issue Date
2025.17.12
Degree
Master of Business Administration
Publisher
Dublin Business School
Rights holder
Rights
Open Access
Abstract
Inflation forecasting is critical for effective economic planning and policy formulation. However, predicting inflation is challenging due to multiple external factors such as housing market trends and immigration. This study explores a hybrid modelling approach to enhance inflation prediction by combining the strengths of traditional statistical methods and advanced predictive techniques. Using time series data for Ireland and the United Kingdom, this research integrates Seasonal Autoregressive Integrated Moving Average with Exogenous Variables (SARIMAX) with other models, including Random Forest, Support Vector Regression, and Long-Short-Term Memory. Four hybrid models SARI-SVR, SARI-RF, RF-SVR and SARI-LSTM were developed and evaluated based on their ability to capture linear patterns and complex nonlinear relationships in the data. The findings demonstrate that the SARI-LSTM model consistently outperformed the others, achieving the lowest Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) for Ireland and the UK. This model’s ability to combine SARIMAX’s seasonal trend analysis with LSTM’s strength in handling sequential dependencies makes it particularly effective for inflation forecasting. By leveraging hybrid modelling, this study provides a comprehensive framework for addressing the complexities of inflation prediction. The results highlight the potential for improved forecasting accuracy, offering valuable insights for policymakers and economists.