Support Vector Machine for Stroke Risk Prediction

Abstract

Nowadays, there are many paralytic patients sent to the hospital, and it is impossible for hospitals to search out those particular patients at once because these patients share the same symptoms such as unconsciousness as those hypoglycemic patients. Nowadays, most patients need to conduct blood tests to ensure whether they have a stroke or not. However, this will take a lot of time and resources. The investigation aims at stroke prediction. By using machine learning, the model will be used to train and test data. Support vector machine (SVM) will be used in this project. By using a support vector classifier (SVC), the model will be trained to learn from data. Then it will react to another data set to find out if it is fitted. As a classification problem, the accuracy is 0.77. It shows that certain model performs well after training which reflects that the prediction is successful. What’s more, its high recall which is 0.83 means that the model of stroke prediction can surely offer some help to patient classification to some extent because it can find most of the paralytic patients among all the samplers. Stroke prediction trained by the SVM model can help make the first division among all patients which can help save a lot of time and energy. However, since there is still a little deviation, it is still need to keep pace with modern medical technology to improve it.