Predictive Vacancy Rate Forecasting ML Pipeline

Design a comprehensive Python machine learning pipeline using scikit-learn and pandas that predicts property vacancy rates for multi-unit residential complexes. The model should incorporate historical occupancy data, neighborhood economic indicators, seasonal trends, and local development metrics. Implement cross-validation with at least three regression algorithms, generating a comparative performance report with RMSE and R-squared metrics. Include a Flask API endpoint that allows real-time prediction inputs and returns probability-weighted vacancy forecasts.