Enhancing Machine Learning Model Interpretability in Intrusion Detection Systems through SHAP Explanations and LLM-Generated Descriptions.