Time Series Forecasting using Hybrid Holt’s ESM with Damping Parameter and Neural Network Model

Time series forecasting has gained much importance in various fields, and the choice of an appropriate model for forecasting is still a significant challenge. Holt's ESM with damping trend is widely used for modeling linear trends, and ANNs are effective in capturing complex nonlinear relationships. However, limitations arise when these models are implemented separately. Hybrid models in time series forecasting are used to combine strengths of different models to cover their shortcomings. In this paper, we develop a hybrid forecasting model that combines Holt's ESM with damping trend and Artificial Neural Network. Holt's ESM with damping trend captures the linearity in the time series, while the non-linearity in the residuals is captured by ANN. Results based on simulation and real-world data show that the hybrid model outperforms the component models. Keywords: ESM, Damping trend, Artificial neural networks, Time series forecasting, Hybrid forecast