Robust Time Series Anomaly Detection Pipeline

Create an end-to-end anomaly detection framework capable of handling multivariate time series data with complex seasonal patterns and irregular intervals. Implement ensemble methods combining statistical techniques (ARIMA, Prophet), machine learning models (Isolation Forest, Local Outlier Factor), and deep learning approaches (LSTM autoencoders). Develop a flexible architecture that can automatically adapt to different data characteristics, with configurable sensitivity thresholds and automated reporting mechanisms. Include comprehensive error handling, performance logging, and a mechanism for human-in-the-loop validation.