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

34 CHAPTER 3. EXPERIMENTS Model inputs recall in % precision in % balanced acc in% accuracy in % TP FP FN TN Oper Rules from 2004 35 27 60 84 38 104 71 872 3-class DNN rain-10 29-35 14-29 47 73 2-class DNN rain-10 42 28 59 78 60 158 82 780 2-class DNN rain-10 nb-10 54 25 64.5 72 77 237 65 701 2-class CNN rain-10 44 32 65 80 62 133 80 805 Table 3.4: Best Performance in each type of Neural Networks (DNN, CNN 2 or 3 classes) compared to Operational Rules (TP, FP, FN, TN, standing for True Positive, False Positive, False Negative, True Negative)