Comparing Logistic Regression and Neural Networks for Clinical Stroke Risk Prediction
An applied comparison of three machine-learning models for predicting stroke from clinical and social risk factors: logistic regression, a dense neural network, and a convolutional neural network, trained on a public ~5,000-record dataset with a binary stroke outcome. Because false negatives are especially dangerous in health care, the models are tuned and compared with an emphasis on minimizing missed cases rather than maximizing overall accuracy, weighing accuracy against computational cost.
2501.00048
This applied study compares machine-learning models for predicting whether an individual will suffer a stroke from clinical and social features (age, hypertension, heart disease, average glucose leve…