Aplikace ANFIS modelu vysvětlitelné umělé inteligence pro podporu rozhodování v chytrých městech

Abstract

This thesis examines the use of the ANFIS model as an explainable artificial intelligence tool for decision support in smart cities. It focuses on the area of air quality monitoring and assessment, specifically on concentrations of PM2,5 fine particulate matter, which is a significant factor affecting public health. The thesis first defines the theoretical and methodological foundations of fuzzy logic, fuzzy inference systems, and the ANFIS model, and then addresses the current use of artificial intelligence in smart cities and the issue of data-driven decision-making. The practical part focuses on the design and creation of ANFIS models using real-world data, incorporating meteorological variables, temporal characteristics, and lagged PM2,5 values. The results show that the inclusion of lagged PM2,5 values significantly improves the model’s predictive ability and that, during design, it is necessary to find a compromise between accuracy and interpretability. ANFIS thus appears to be a suitable approach for decision support in the field of urban air quality.