RINA - Démonstrateur d’utilisation de l’intelligence artificielle pour la gestion opérationnelle des risques naturels gravitaires géologiques | FEREC 2020

Random Forest actual 0 1 predicted 0 849 94 1 95 52 Figure 2 : confusion matrix for Random forest model adopting a weight 3 times greater for predicting there will be an event Random Forest Operational accuracy 0.80 0.84 balanced accuracy 0,62 0.60 precision 0,35 0.35 recall 0,34 0.27 F1 0,35 0.30 Table 1: indices for the Random forest model and the Operational model Additional result of models : Histogram based gradient boosting Simple model made with no optimisation. The default value, defined by sklearn, was used for most of the hyper-parameters with the exception of the maximum depth that was fixed at 3. HistGradBoosting actual 0 1 predicted 0 901 109 1 43 37 Figure 1 : confusion matrix for histogram based gradient boosting model 52