Comparison of outcomes classical and novel forecasting tools

To evaluate the predictive power of various competing forecast models, we use two commonly applied metrics for evaluating forecasts, i.e. Root Mean Squared Error (RMSE), the Mean Absolute Error (MAE) 

with y,f, i,t the forecast of  yi,t and  h  the forecast horizon

Comparison of RMSE and MAE for various models


RMSEMAE

ME model with country-specific intercepts and general trend

14.349.54

ME model with country-specific intercepts and trends

3.791.95
ME model with country-specific intercepts and general trend, squared trend and lagged acreages4.753.51
MERF model with country-specific intercepts and trend in RF14.306.98
MERF model with country-specific intercepts and trend, squared trend and lagged acreages in RF2.851.32
Random Forest (RF) model with country-specific intercepts and trends, general squared trend and lagged acreages3.861.72




Besides these metrics we can also visually compare the forecasts of various method.


The combination of a classic Mixed Effects model with a machine learning Random Forest performs best, since it combines the strength of the ME model to exploit the country specificity, with the superior forecasting capability of the random forest.






Root Mean Squared Error (RMSE)

Mean Absolute Error (MAE)

Comparison of RMSE and MAE for various models and countries

11/12/2024 08:51:05
  • Cover
  • Intro Playbook
  • The biomass acreage problem
  • Classical forecasting tools
  • Novel forecasting tools
  • Hemp and Flax case
  • Socio-economic context
  • Socio-economic context2
  • Comparison of outcomes (copy 2)
  • Outlook
  • References
  • Colofon and Contact
04/16/2024 00:00:00