An adaptive control system with a dynamic architecture was developed and evaluated. The system integrates heterogeneous sensor networks with machine learning models to enhance occupant comfort and energy efficiency. The dynamic architecture combines edge-cloud data processing with predictive algorithms, enabling real-time monitoring of environmental and behavioral parameters. Performance evaluation results showed that the occupant comfort index was consistently maintained within a narrow band between indices 26 and 32, with a sharp peak at 29. Correlation analysis results showed a positive relationship between illuminance and comfort [Pearson correlation coefficient (r) = +0.73], and moderate and weak negative correlations with relative humidity (r = −0.57) and ambient temperature (r = −0.36). Despite high-fidelity data acquisition, predictive modeling showed prediction flattening, with random forest regression outputs compressed between 26.5 and 27.5, failing to capture dynamic fluctuations in occupant perception. This limitation underscores the need for advanced sensor technologies capable of integrating physiological signals, such as heart rate variability and localized thermal dissipation, to overcome environmental inertia. The results of this study underscore the importance of continuous model training, multimodal sensor fusion, and feedback mechanisms in advancing adaptive architecture toward personalized, occupant-centric management.