Fault analysis, machine learning, model interpretability, SHAP, telecommunications