Asthma is a chronic respiratory disease marked by dynamic symptom variability and strong sensitivity to environmental and behavioral factors, necessitating continuous monitoring and early intervention. Traditional asthma management approaches rely on periodic clinical visits and self-reported symptoms, which often fail to capture real-time physiological changes and environmental exposure. This paper presents ADAPT-AsthmaNet, an adaptive Internet of Things enabled framework that integrates multimodal sensing with deep learning and reinforcement learning for real-time asthma exacerbation prediction. The system continuously collects physiological and environmental data using wearable and portable IoT sensors and processes these signals within a cloud-based analytics platform. A hybrid deep learning architecture combining convolutional neural networks, Long Short-Term Memory networks, and an attention mechanism is employed to model both short-term signal variations and long-term temporal dependencies. A Proximal Policy Optimization–based reinforcement learning module further enables personalized and adaptive alerting. Evaluation using the AAMOS-00 real-world dataset demonstrates that ADAPT-AsthmaNet achieves superior performance, attaining 94.1% accuracy with high sensitivity and specificity. The results highlight the potential of adaptive IoT-driven intelligence for personalized asthma monitoring and proactive care