DEVELOPMENT OF A CLOUD SAAS PLATFORM BASED ON MACHINE LEARNING FOR MONITORING AND FORECASTING ATMOSPHERIC AIR QUALITY IN CITIES

Abstract

In this paper, the architecture of a computing service of the "SaaS" type is proposed for the implementation of numerical modeling based on an improved fractional-differential stochastic model. The main purpose of the research is to develop a modular system that allows users to perform high-precision calculations without the need to use their own computing resources.

The proposed solution consists of several interconnected subsystems: data reception, queuing and preprocessing tasks, a mathematical calculation module, analysis and visualization of results, system monitoring, logging and security. Each subsystem implemented using modern software tools (Julia, Python, Django, Prometheus, etc.) interacts via APIs and relies on a centralized PostgreSQL-based data warehouse.

As a result, a scalable and fault-tolerant architecture was created that provides automatic acceptance of requests, calculations, and interpretation of the results. In conclusion, it is emphasized that such a distributed approach is effective and can be adapted to other tasks of scientific modeling and analytics.