Fig. 4: Box-and-whisker plots showing performance of the 2400 \({{{{{{{{\mathcal{M}}}}}}}}}_{i}^{\ell }\) models of set \({{{{{{{\mathcal{O}}}}}}}}\), in the model testing phase. | Nature Communications

Fig. 4: Box-and-whisker plots showing performance of the 2400 \({{{{{{{{\mathcal{M}}}}}}}}}_{i}^{\ell }\) models of set \({{{{{{{\mathcal{O}}}}}}}}\), in the model testing phase.

From: Deep learning forecast of rainfall-induced shallow landslides

Fig. 4

For each lag, ℓ (x-axis), purple boxplots show variability of the AROC of the 100 \({{{{{{{{\mathcal{M}}}}}}}}}^{\ell }\) models (left y-axis), and violet boxplots show Balanced Accuracy, BA51 (right y-axis). White horizontal lines show median value for each boxplot. Boxplot legend: min and max are minimum and maximum values; 25th, 50th (median), and 75th are percentiles.

Back to article page