Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{201193,
        author = {Ganga Vijay Kumar and B.Pramodhini},
        title = {A Real-Time Monitoring Framework for Deployed Machine Learning Models with Drift Detection and Performance Analytics},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {2885-2890},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=201193},
        abstract = {Machine learning models have been broadly de-ployed in real world application, for example, making predictive decisions.However, performance of the deployed model degrades rapidly with time since data distribution will change. The phenomenon is widely known as data drift or concept drift and cause problems for a deployed model. In this paper, we introduce a monitoring framework designed for the real time evaluation and sustainment of a deployed machine learning model in a changing world.A customer churn dataset which contains 7032 samples and 31 features was utilized to train a set of machine learning models, which include Logistic Regression, Random Forest and XGBoost. And they are split into train set (80%) and test set (20%). Random Forest is chosen among them as the final model, which has the accuracy of 79.45% and is deployed in a FastAPI based service to mimic real time prediction. Our system continuingly record the prediction logs, evaluates and monitors the performance trend of the model and employs PSI (Population Stability Index) to capture distribution drift for features. An interactive dash board with Streamlit is also provided to visualize the behavior, performance and drift features of model. The result from experimentation confirms that our system can effectively detect the distribution shift of features (e.g. Tenure and billing attributes), which represents the possibility of model performance degradation at the early stage. This work focuses on post-deployment monitoring system and can be a practical approach for a real world application to guarantee a sustainable performance for machine learning models.},
        keywords = {Machine Learning, Model Monitoring, Data Drift, Population Stability Index, Random Forest, Real-Time Systems},
        month = {May},
        }