Using Machine Learning Algorithm as a Method for Improving Stroke Prediction

Having sudden strokes has had a very negative impact on all aspects in society to the point that it attracted efforts for better improvement and management of stroke diagnosis. Technological advancement also had an impact on the medical field such that nowadays caregivers have better options for taking care of their patients by mining and archiving their medical records for ease of retrieval. Furthermore, it is quite essential to understand the risk factors that make a patient more susceptible to strokes, thus there are some factors that make stroke prediction much easier. This research offers an analysis of the factors that enhance the stroke prediction process based on electronic health records. The most important factors for stroke prediction will be identified using statistical methods and Principal Component Analysis (PCA). It has been found that the most critical factors affecting stroke prediction are the age, average glucose level, heart disease, and hypertension. A balanced dataset is used for the model evaluation which was created by sub-sampling since the dataset for stroke occurrence is already highly imbalanced. In this study, seven different machine learning algorithms are implemented: Naïve Bayes, SVM, Random Forest, KNN, Decision Tree, Stacking, and majority voting to train on the Kaggle dataset to predict occurrence of stroke in patients. After preprocessing and splitting the dataset into training and testing sub-datasets, these proposed algorithms were evaluated according to accuracy, f1 score, recall value, and precision value. The NB classifier achieved the lowest accuracy level (86%), whereas the rest of the algorithms achieved similar accuracies 96%, f1 scores 0.98, precision 0.97, and recall 1.