Machine Learning for Asthma Prediction: Analyzing Health Data for Early Detection

Asthma is a persistent respiratory condition that significantly affects people worldwide, making the prediction of asthma flare-ups crucial for enhancing patient care and optimizing healthcare resources. Despite progress in the field of machine learning, current predictive models often face challenges due to their dependence on historical clinical data alone. This paper evaluates existing asthma prediction models and introduces a new approach utilizing Bernoulli distribution and logistic regression to enhance prediction accuracy. Current models fre- quently depend on clinical variables such as coexisting conditions, past exacerbations, and healthcare engagement. Widely used, methodologies include decision trees, random forests, and logistic regression; however, these models generally perform poorly when evaluated on varied real-world datasets. A notable drawback is their inability to factor in environmental influences, such as air quality, weather patterns, and pollen counts, all of which play a role in asthma exacerbations. In future research, we will further investigate and refine techniques to better integrate a variety of data sources, aiming to make asthma prediction models more robust, dependable, and suitable for diverse populations.