Student Performance Predictive Model with Feature Engineering

Design a comprehensive machine learning pipeline using pandas and scikit-learn to predict student academic performance. Develop a model that incorporates feature engineering techniques including polynomial features, interaction terms, and time-series based academic trajectory analysis. Include preprocessing steps for handling missing data from multiple school information systems, and create a model that can predict student performance with at least 85% accuracy. Implement cross-validation strategies and generate a detailed model interpretability report using SHAP values.