Sultan, Z. A., & Aljboori, N. S. K. (2023). Seasonal decomposition and trend using hybrid STL-FNN with application. Advances in the Theory of Nonlinear Analysis and Its Application, 7(4), 35–46.
- ISSN: 2587-2648
Menu
Zena A. Sultan, Nihad S. Khalaf Aljboori
Department of Mathematics, College of Education for Women, Tikrit University, Tikrit, Iraq.
In time series analysis, most researchers rely on traditional statistical models; however, recent advances have introduced neural network–based approaches that mimic the functioning of biological neurons. This paper proposes a novel hybrid model, STL–FNN, which integrates Seasonal and Trend Decomposition using Loess (STL) with a Feed-Forward Neural Network (FNN).
The STL method decomposes the original time series data into three subseries: seasonality, trend, and residual components. Each of these components is then modeled using the FNN, which predicts the seasonal component while separating the trend and residual parts. The final predicted outputs are aggregated to produce the overall time series forecast.
The performance of the proposed STL–FNN hybrid model is evaluated using the Mean Absolute Error (MAE) criterion. Real-world time series data—specifically, monthly spending of foreign visitors in the United Kingdom from January 1986 to February 2020—were analyzed to assess the model’s predictive capability. Results confirm that the STL–FNN model achieves superior accuracy and robustness compared to conventional methods.
Keywords: Decomposition; STL model; FNN; Smoother Loess; STL–FNN structure.
Sultan, Z. A., & Aljboori, N. S. K. (2023). Seasonal decomposition and trend using hybrid STL-FNN with application. Advances in the Theory of Nonlinear Analysis and Its Application, 7(4), 35–46.