A Machine Learning Data-Driven Approach for Predicting Cardiovascular Disease

Cardiovascular diseases (CVDs) are prevalent health disorders in the human population. Early detection can lower mortality rates by preventing or minimizing CVDs. One possible approach is to use machine learning algorithms to identify risk variables. In order to accurately forecast cardiac disease, we want to provide a model incorporating many approaches. We have gathered data from Boston University and employed data transformation and pre-processing techniques to produce reliable training model information to run our suggested model properly. The data set comes from the Framingham Heart Study. Three experimental analyses were conducted on the study: All features, selected features by univariate feature selection, and dimensionality reduction in an unsupervised manner using PCA. Several ensembles (boosting, bagging, tailored) and classical (KNN, NB, DT, SVM, LR) models were applied in the first experiment; based on the results of these tests, specific hybrid and selected models were employed in the second and third experiments. The outcome analysis section goes into great detail about it. Holding the overall result performance constant, the Random Forest (RF) yielded consistent outcomes, averaging close to 94%, besides the most dependable and highest result given by GB in the second experiment, which is 98%, with a log loss of almost zero(0.07).