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

Figure 9: mass of those rock falls in fonction of the location and histogram representing their frequencies. 3.3. PCA analysis of rockfall events Présenter données sur lesquelles la PCA est faite? Pourquoi (P0, P1, P2, P3, P4) ? Pourquoi réduire la dimension dans ce cas précis? Principal Component Analysis is a dimension reduction method that creates new features based on the existing ones. It creates the same amount of features as previously existed and by combining the different features is capable of creating new ones that are more adapted to explaining the data in fewer dimensions. The objective is being able to plot and create models on lower dimensions. With the new variables it is important to observe how much of the variance each feature is capable of explaining and consequently how effective the reduction of dimensions is. For the data before and after the implementation of protective structures the figure 10 (a) and (b) was created based on the rain of the previous 5 days (P0, P1, P2, P3, P4). Both bar graphs are extremely similar, with all values being superior to 10% indicating that ideally all features should be considered. The first analysis that should be made is with the two first components. Together they explain 60% of the variance. 17