Architecture of Intelligent Management in Administrative Web Applications Using MLOps and Machine Learning

Abstract

This paper explores the development and implementation of an Intelligent Management System (IMS) 
integrated into an administrative web application, utilizing machine learning and data analysis methods. The relevance 
of the topic is driven by the need to enhance the efficiency of digital management processes through decision-making 
automation and predictive analytics. 
The aim of the study is to build an adaptive system architecture capable of real-time state analysis, risk forecasting, 
and autonomous control actions without human intervention. Methods used include gradient boosting algorithms, 
autoencoders, reinforcement learning, and MLOps tools for model monitoring and automatic retraining. 
A modular architecture was developed, encompassing data collection and processing, training pipelines, REST API, 
and a control unit. The system was tested in both a simulated environment and a real-world web application. Results 
showed a forecast accuracy increase up to 92%, a 5–6 fold reduction in response time to failures, and improved system 
robustness against anomalies. 
The proposed solution demonstrates high practical value and can be adapted to various digital management domains—
 from IT infrastructure to logistics and document workflow. The research was supported by a grant under the scientific 
research program of the Republic of Kazakhstan.