Automated Student Performance Predictive Analytics Pipeline

Design a comprehensive Python data pipeline using pandas and scikit-learn that ingests multi-source student performance data (LMS logs, assessment scores, attendance records) and builds a predictive model forecasting student risk of academic failure. The solution must include automated feature engineering, handle missing data dynamically, implement at least three machine learning algorithms (logistic regression, random forest, gradient boosting), and generate an interpretable risk dashboard with confidence intervals. Include robust error handling and logging mechanisms compatible with enterprise-scale educational data systems.