Fig. 9
From: ProWaste for proactive urban waste management using IoT and machine learning
SHAP-based feature-importance analysis of the decisive model. (a) Mean absolute SHAP values per class, confirming dominance of the three selected features. (b) Global mean SHAP bar chart: ‘TimeSinceLastMaintenance_days‘ is by far the most influential predictor. (c) SHAP summary-dot plot showing the distribution of impacts; positive values shift predictions toward the High class. (d)–(f) Waterfall plots, illustrating how the key features increase or decrease the predicted criticality score of a particular instance (first sample).
