Improving ADHD Detection with Cost-Sensitive LightGBM

One common neurodevelopmental issue that needs prompt and precise diagnosis for successful treatment is attention-deficit/hyperactivity disorder (ADHD). In this study, we present a novel approach using a cost-sensitive LightGBM (Light Gradient Boosting Machine) model to enhance the prediction of ADHD. Our key contribution lies in the integration of cost-sensitive learning, which prioritizes minimizing the misclassification costs associated with false negatives and false positives, critical in medical diagnostics. Additionally, we leverage the computational efficiency and speed of LightGBM, a gradient boosting framework known for its high performance in handling complex data structures. Our results demonstrate that the cost-sensitive LightGBM model achieves a predictive accuracy of 94% and an F1-score of 94%, while significantly reducing the time required for training and inference. This makes it a viable tool for real-world ADHD screening applications. This study highlights the importance of incorporating cost-sensitive algorithms in medical predictions to achieve a balanced trade-off between accuracy and resource efficiency.