Model-agnostic explainable artificial intelligence tools for severity prediction and symptom analysis on Indian...

Area of Science:

• Artificial Intelligence in Healthcare
• Machine Learning Interpretability
• COVID-19 Research

Background:

• The COVID-19 pandemic necessitated advanced AI solutions for resource management and risk identification.
• Many AI models lacked practical applicability due to their "black-box" nature, hindering interpretability.
• Explainable Artificial Intelligence (XAI) emerged to address the interpretability challenge in machine learning models.

Purpose of the Study:

• To explore the application of model-agnostic XAI techniques for COVID-19 symptom analysis.
• To develop and evaluate machine learning models for COVID-19 severity prediction in Indian patients.
• To assess the interpretability and trustworthiness of AI models using XAI tools.

Main Methods:

• Leveraged machine learning models including Decision Tree Classifier, XGBoost Classifier, and Neural Network Classifier.
• Employed model-agnostic XAI methods: SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME).
• Conducted COVID-19 symptom analysis and severity prediction tasks on Indian patient data.

Main Results:

• XAI tools successfully augmented AI system performance with human-interpretable evidence.
• Interpretability plots demonstrated the reasoning behind model predictions.
• Comparative analysis highlighted the significance and impact of XAI in a healthcare context.

Conclusions:

• SHAP and LIME analysis are promising for developing interpretable AI models in healthcare.
• XAI enhances the trustworthiness and practical applicability of AI systems.
• The study advocates for the integration of XAI in future machine learning model development for better healthcare outcomes.