Figure 3
Forecasting the prevalence of people with insufficient food consumption in Yemen. The forecasting is performed over 5 different monthly splits of all governorates time series, from October 2021 to February 2022. (a) Box plots of the coefficient of determinations (\(R^2\)) across the 5 splits for both the proposed and the naive models (in blue and orange, respectively), for each forecasting horizon. (b) Box plots of the mean squared error (MSE) across the 5 splits for both the proposed and the naive models for each forecasting horizon. (c) Box plots of the difference between the \(R^2\) of the proposed and of the naive model for each split. (d) Box plots of the difference between the MSE of the naive and of the proposed model for each split. (e) Predicted vs actual value for each data point in the 5 splits. Colors represent the corresponding forecasting horizon and vary from dark blue (1 day) to yellow (30 days). (f) Example of forecasting results for December 2021 in Amanat Al Asimah (top) and February 2022 in Abyan (bottom).
