Predicting Asthma Risk using Internet of Things and Machine Learning

Asthma affects approximately 262 million people globally, with air pollution recognized as a leading trigger for asthma attacks. Yet, air quality monitoring tools are often costly and inaccessible, forcing asthma patients to rely on general weather data instead of precise, location-based insights. This paper proposes an accessible Internet of Things (IoT) solution combined with a machine learning-based prediction algorithm to assess individual asthma risk in real-time. The proposed IoT device integrates microcontrollers and sensors to capture environmental data near the user, including temperature, humidity, hazardous gas concentrations, and particulate matter levels. Additionally, health indicators like heart rate and oxygen levels may be monitored for a more comprehensive assessment. By utilizing machine learning algorithms to analyze this data against established high-risk air quality conditions, the system generates a personalized asthma risk prediction. Data is periodically uploaded to a web application or database, where real-time risk predictions and historical data remain accessible to the user. This work is novel in its use of a dual IoT system for combining both environmental and physiological data, enabling a more holistic risk prediction. Machine learning models including Random Forest and Support Vector Machines were evaluated, with Random Forest achieving 92% accuracy. Through testing across various locations and participants, we validate the system’s accuracy and reliability, demonstrating the potential of IoT and machine learning in affordable, individualized asthma management.