Automated Machine Learning Model Management

We present an automated machine learning model management system, which provides feedback loops, at-rest and in-flight observability, and continuous monitoring for sustaining and improving model performance in the long term. The system is built to support automated retraining and redeployment to mitigate challenges associated with data drift or the evolution of operational norms. Our models achieved 99.14% accuracy, 99.19% precision, 99.17% recall, and 99.18% F1 scores in Random Forest and Gradient Boosting. Our system demonstrated superior on-line observability and adaptability to AutoWeka and more conventional Gradient Boosting. Continuous feedback with on-going observations allows users to leverage our system in scalable, decision making applications, in non-stationary conditions while ensuring ongoing performance.