Student Performance Predictive Model with Early Warning System

Design a comprehensive machine learning pipeline in Python that predicts student at-risk of academic failure using multi-dimensional data sources. Integrate data from learning management systems, attendance records, assignment submissions, and historical academic performance. Develop a predictive model using scikit-learn that generates a risk probability score, with an emphasis on interpretable features. Create an automated alert system that flags students with >60% likelihood of academic underperformance, including recommended intervention strategies.