Machine learning models for preventative mobile health asthma control.

Area of Science:

• Computational health informatics
• Machine learning applications in respiratory medicine

Background:

• Asthma attacks are frequently triggered by environmental pollutants, viruses, physical activity, and allergens.
• Predictive modeling offers a potential strategy for proactive asthma management.

Purpose of the Study:

• To develop a machine learning model for predicting asthma attack triggers using mobile health technology data.
• To provide timely warnings to patients, enabling them to avoid specific triggers.

Main Methods:

• Lightweight machine learning models (XGBoost, Random Forest, LightGBM) were trained and evaluated.
• A 70-30 train-test split was employed, with performance assessed using Precision, Accuracy, Recall, F1 scores, and model speed.

Main Results:

• The XGBoost model achieved high performance with an Accuracy of 0.902, Recall of 0.904, and F1 score of 0.860.
• The model demonstrated a training speed of 14 seconds, indicating efficiency.

Conclusions:

• Predicting asthma triggers via machine learning is a viable approach for asthma control.
• Actionable trigger information empowers patients to modify behaviors, though further mobile health system integration is needed.