Enterprise Financial Anomaly Detection System

Design an unsupervised machine learning framework for detecting financial anomalies across complex organizational financial transactions. Develop a multi-stage approach using isolation forests, autoencoders, and Gaussian mixture models that can identify statistically unusual patterns in general ledger entries, inter-company transfers, and expense categorizations. Include a robust explainability module that provides contextual reasoning for each detected anomaly.