Use Feed-Forward Neural Network to Predict the Brain Stroke Based on One-Dimensional Data

World Health Organization (WHO) has classified stroke as a dangerous disease. Currently, there are no effective methods to treat strokes, but if strokes could be accurately predicted in advance, the disability and death will decrease. Traditional stroke diagnosis heavily relies on the clinical experience of doctors, and the diagnosis process consumes time and energy. Machine learning has developed in recent years, which makes it possible to predict strokes. This article uses Feed-Forward Neural Network (FNN) and traditional machine learning methods, which include Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), and decision tree (DT) methods, then compares the prediction performance among these methods. The FNN achieved the best performance, which reaches an accuracy rate of 81% at most, and the accuracy rate is very close to methods that use other methods. One-dimensional data simplifies the hardness of the experiment, it lowers the cost of prediction, and the accuracy could be further improved if the dataset includes continuous data such as heart rate and blood pressure, and these items could preserve timing information. This article aims to provide some support for researchers working in related fields.