Fig. 2: Machine learning model performance in predicting system size over varying observation periods.
From: How key features of early development shape deep convective systems
A Evolution of Pearson-r regression score in green and mean square error in blue for the estimate of maximum extension \({{\mathcal{L}}}_{max}\) in the three machine learning models – Linear Regularized Regression in solid line, Multi Layer Perceptron in dash-dotted line and Random Forest in dashed line – trained on the evolution of the growth rate during a considered observed period. B Same with physical features considered, which include system shape, physical field mean and standard deviation, trajectory and neighbor influence. The Multi Linear model performances from the top panel are repeated in the lower panel in gray to ease comparison (ascending is pearson-r, descending is rmse). The horizontal blue dotted line indicates the standard deviation of the dataset, representing the performance of a random prediction.
