DSpace Repository

Machine Learning Based Approach for Stroke Prediction

Show simple item record

dc.contributor.author Afroze, Minara
dc.date.accessioned 2023-01-15T09:00:22Z
dc.date.available 2023-01-15T09:00:22Z
dc.date.issued 22-11-14
dc.identifier.uri http://dspace.daffodilvarsity.edu.bd:8080/handle/123456789/9340
dc.description.abstract When a blood vessel that supplies the brain with oxygen and nutrients becomes blocked by a clot or breaks, a stroke happens. Nowadays, it is the most common cause of death in the entire world. Every four minutes, someone become a victim of stroke and passes away, although 80% of stroke deaths may be avoided if we could recognize or anticipate them before they happened. Early stroke detection may be preferable to reducing the severity of the condition. Data science has played a significant role in the development of medical studies in recent years. To predict the chance of a stroke, several machine learning approaches are developed that use a patient's physical and physiological reporting data. The most significant risk factors for stroke in patients include age, cardiac disease, average blood sugar level, and hypertension. In this study, we employ Decision Tree, XG Boost, Light Gradient Boosting Machine (LGBM), Random Forest and K-nearest Neighbors learning as five machine learning algorithms to determine the most accurate model that can anticipate the risk of stroke and the dataset was collected through Kaggle. In this study, comparing with other machine learning algorithms utilized, the testing results indicate that the Random Forest algorithm has the maximum accuracy rate that is 96%. Keywords: Ischemic stroke, Hemorrhagic stroke, Light Gradient Boosting Machine, Precision-Recall Curve · Random Forest · ROC Curve· Stroke Prediction. en_US
dc.language.iso en_US en_US
dc.publisher Daffodil International University en_US
dc.subject Machine learning en_US
dc.subject Algorithms en_US
dc.title Machine Learning Based Approach for Stroke Prediction en_US
dc.type Other en_US

Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


My Account