Scalable Brain Stroke Prediction Using PySpark: A Big Data Approach

Brain stroke affects a large number of the population and is one of the primary causes of death and permanent disability, therefore the identification of risk factors is paramount for timely prevention. These interactions are difficult to capture because stroke risk factors are demographics, life style and medical history all of which cannot be modeled using statistical solution. To tackle these and similar issues, the present paper employs PySpark, a distributed computing system, for the formulation of small-scale and big data with various machine learning algorithms including, Logistic Regression, Gradient Boosting, Decision Trees, Naive Bayes, and Random Forests, for predicting brain stroke. The used models were derived from a great amount of healthcare data, though the hyperparameter optimization was applied to improve the results. Among those algorithms, Gradient Boosting showed the best result with the accuracy of 92.2, Random Forest is 90.5, which indicates that these algorithms can give good results at the time of risk assessment of stroke. Measuring techniques like precision, recall and F1 score were employed to measure model accuracy and test ability to control between False Positives and False Negatives. This multiple model approach recognizes the most efficient algorithms for possible implementation in real-life, mass, clinical practice for early detection, patient control, and individual prevention.