A Review and Performance Study of AI Techniques in Asthma Prediction

Abstract

Asthma is a chronic respiratory disease that affects millions of people worldwide and continues to impose a significant burden on healthcare systems due to frequent exacerbations, hospitalizations and treatment costs. Recent advances in Artificial Intelligence (AI) have made the development of intelligent prediction models possible for early diagnosis, risk assessment and personalized disease management. This review discusses recent studies for AI-based asthma prediction and role of structured and unstructured clinical data, Machine Learning (ML), Deep Learning (DL) and Natural Language Processing (NLP) techniques. Some of the commonly used algorithms include Random Forest, XGBoost, Support Vector Machines, Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks and transformer-based NLP models. The studies reviewed show that ensemble learning and deep learning approaches tend to outperform the traditional statistical methods in terms of prediction. Moreover, intelligent healthcare systems have growing potential as manifested by AI applications such as smart inhalers, EHR-based symptom extraction and non-invasive breath analysis. However, promising results are still hindered by important challenges such as data imbalance, model interpretability, privacy, and dataset standardization. Future work should focus on developing explainable, privacy-preserving, and clinically deployable AI frameworks for the prediction of asthma.