Fig. 9: Feature importance in the conventional ML experiment that provided the best accuracy.
From: AI perceives like a local: predicting citizen deprivation perception using satellite imagery
This experiment involved a Random Forest (RF) model using the full set of log-transformed features. The colour indicates the dataset from which the features were derived. The features that have the strongest influence on the predicted deprivation score were also identified by citizen scientists as having a strong influence on their perception of deprivation (e.g., street network quality, built-up fabric density, presence of open spaces, etc.).
