Impact of Coiflet Wavelet Decomposition on Forecasting Accuracy: Shifts in ARIMA and Exponential Smoothing Performance

Abstract

Accurate forecasting of electricity demand is crucial for effectiveenergy management and resource planning. Recently, time seriesforecasting methods such as Exponential Smoothing (ES) andARIMA models have gained popularity due to their ability to detectintricate seasonal patterns in data. This study examines howvarious wavelet families, particularly the Coiflet wavelet, impact theperformance of ES and ARIMA models in forecasting electricitydemand. We observed that applying the Coiflet wavelet couldsignificantly enhance forecasting outcomes. The study evaluates theeffectiveness of different wavelets in improving the forecastingaccuracy of ES and ARIMA models, with a particular focus on thesuperior performance of Coiflet wavelets. Our findings offerinsights into the suitability of wavelet-based methods for electricitydemand forecasting. Nonetheless, the choice between ARIMA andExponential Smoothing should be guided by the specificcharacteristics of the time series data and forecasting goals. Forcomplex and noisy data, ARIMA combined with Coiflet waveletpreprocessing proves to be a robust and effective forecastingapproach, demonstrating superior performance in our analysis.