Scholarlink

<p>Accurate inflation forecasting is critical for effective economic planning and monetary policy formulation, especially in emerging economies such as Nigeria. Persistent inflation volatility driven by structural inefficiencies, external shocks, and weak monetary frameworks continues to challenge policy responses and macroeconomic stability. Traditional econometric models like Seasonal Autoregressive Integrated Moving Average (SARIMA) often fail to capture the nonlinear dynamics inherent in inflation behavior, while standalone machine learning models such as Artificial Neural Networks (ANNs) lack transparency and reliability in highly variable economic settings. To address this gap, this study proposes a hybrid forecasting framework that integrates SARIMA and Artificial Neural Network models to leverage their respective strengths in modeling linear and nonlinear patterns in Nigeria’s inflation rate. The methodology involves fitting a SARIMA (2,1,2)x(1,1,0)<sub>12</sub> model to the historical training set (monthly inflation data from January 2003 to September 2017) to capture trend and seasonality. Residuals and SARIMA-fitted values were then used alongside lagged inflation variables as input features for training a Multi-Layer Perceptron ANN model. A grid search with time-series cross-validation was employed for hyperparameter tuning, ensuring optimal model performance. The final hybrid model was used to forecast inflation rates for the period March 2024 to December 2030. Results show that the hybrid model outperformed both standalone SARIMA and ANN models across multiple metrics, including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared. The hybrid model achieved a forecast R-squared of 92.09%, with a MAPE of 7.68%, indicating excellent predictive accuracy on the test set. The forecast suggests periods of inflation spikes, particularly in 2025 and 2026, followed by relative stabilization toward 2030. The study concludes that the hybrid SARIMA–ANN model is a powerful and reliable tool for inflation forecasting in Nigeria. It is recommended for policymakers and economic analysts seeking to anticipate inflationary trends and design timely interventions. Future research could incorporate exogenous macroeconomic variables and explore deep learning enhancements to further improve forecasting performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Hybrid Model of Artificial Neural Network and SARIMA Models for Predicting Inflation Rate Change in Nigeria's Economy

  • 期刊论文

  • 发表日期：

    2025年10月18日

查看全文
  • D. C. Bartholomew,
  • H. C. Iwu,
  • I. P. Ibemere

摘要

Accurate inflation forecasting is critical for effective economic planning and monetary policy formulation, especially in emerging economies such as Nigeria. Persistent inflation volatility driven by structural inefficiencies, external shocks, and weak monetary frameworks continues to challenge policy responses and macroeconomic stability. Traditional econometric models like Seasonal Autoregressive Integrated Moving Average (SARIMA) often fail to capture the nonlinear dynamics inherent in inflation behavior, while standalone machine learning models such as Artificial Neural Networks (ANNs) lack transparency and reliability in highly variable economic settings. To address this gap, this study proposes a hybrid forecasting framework that integrates SARIMA and Artificial Neural Network models to leverage their respective strengths in modeling linear and nonlinear patterns in Nigeria’s inflation rate. The methodology involves fitting a SARIMA (2,1,2)x(1,1,0)12 model to the historical training set (monthly inflation data from January 2003 to September 2017) to capture trend and seasonality. Residuals and SARIMA-fitted values were then used alongside lagged inflation variables as input features for training a Multi-Layer Perceptron ANN model. A grid search with time-series cross-validation was employed for hyperparameter tuning, ensuring optimal model performance. The final hybrid model was used to forecast inflation rates for the period March 2024 to December 2030. Results show that the hybrid model outperformed both standalone SARIMA and ANN models across multiple metrics, including Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared. The hybrid model achieved a forecast R-squared of 92.09%, with a MAPE of 7.68%, indicating excellent predictive accuracy on the test set. The forecast suggests periods of inflation spikes, particularly in 2025 and 2026, followed by relative stabilization toward 2030. The study concludes that the hybrid SARIMA–ANN model is a powerful and reliable tool for inflation forecasting in Nigeria. It is recommended for policymakers and economic analysts seeking to anticipate inflationary trends and design timely interventions. Future research could incorporate exogenous macroeconomic variables and explore deep learning enhancements to further improve forecasting performance.