Impact of computing platforms on classifier performance in heart disease prediction

Prediction and classification, a supervised learning technique in machine learning, addresses various challenges related to finding useful patterns present in data. This work explores how different computing platforms influence the accuracy of classification results when employing the same models. Heart disease, a widespread global health issue affecting both men and women, results from a complex interplay of lifestyle factors and genetics. Through visual representations, we examined the diverse factors influencing heart stroke occurrences. We employed multiple classification methods such as Logistic Regression, K Nearest Neighbour (KNN), Support Vector Machine (SVM), Naïve Bayes, and Decision Tree (DT), assessing their accuracy using WEKA and Google Colab (using Scikit-Learn library). Our evaluation revealed that SVM achieves 77% accuracy when implemented using Scikit-Learn, demonstrating superiority over other methods. However, when using WEKA, both logistic regression and SVM demonstrated nearly 91% accuracy using the exact same hyperparameters. This research demonstrated the significance of platform selection in influencing classifier performance, offering valuable insights on how results reported in research can be impacted by the selection of the software and tools, using heart disease prediction as a use case scenario.