Publication

Abstract : Machine learning lifecycle management is a cyclic process followed by most data science projects. It involves various steps, from understanding the business objectives to monitoring the model once deployed. Various open-source tools are present to maintain the accuracy of the deployed model by performing experiment tracking, model monitoring, feature storing, model registry, etc. Most open-source tools focus on completing a particular task at a time. Integrating all these tools into a single framework increases productivity and scalability. It reduces the cost for the organization. This paper provides a method to integrate open-source tools for model monitoring, feature storing, and machine learning experiment tracking operations into a single framework. All these tools are integrated into a single framework in MLflow. The results obtained for the integrated framework are coded in Python using Jupyter Notebook.