Forecasting e-waste recovery scale driven by seasonal data characteristics: A decomposition-ensemble approach

Area of Science:

• Environmental Science
• Data Science
• Circular Economy

Background:

• Accurate forecasting of e-waste recycling scale is essential for effective circular economy policy development and resource management.
• Traditional single forecasting models struggle with the seasonal characteristics of quarterly e-waste data, leading to errors and inconsistencies.

Purpose of the Study:

• To propose a novel forecasting framework, CH-X12/STL-X, for e-waste recycling scale prediction.
• To address the limitations of single models in handling seasonal time-series data.
• To improve the accuracy and stability of e-waste recycling scale forecasts.

Main Methods:

• Utilizing the Canova-Hansen (CH) test to identify seasonal characteristics in e-waste recycling time-series data.
• Applying X12 or Seasonal-Trend Decomposition using LOESS (STL) for seasonal decomposition.
• Employing the Holt-Winters model for seasonal component prediction and Support Vector Regression (SVR) for other components.
• Integrating component predictions using a linear sum for the final forecast.

Main Results:

• The CH-X12/STL-X framework demonstrates superior and more stable forecasting performance compared to traditional single models.
• The empirical results validate the framework's ability to handle diverse seasonal data characteristics.
• The decomposition-integration approach effectively mitigates forecasting errors associated with seasonality.

Conclusions:

• The CH-X12/STL-X framework offers a robust solution for e-waste recycling scale forecasting.
• This approach enhances the reliability of forecasts, supporting better policy-making and resource optimization in the circular economy.
• The study highlights the importance of considering data seasonality in time-series forecasting for environmental and economic applications.