Comparative Analysis of Machine Learning and Hybrid Models for Predictive Modeling

Cardiovascular disease is still a dominant cause of death worldwide. Timely intervention and treatment demand early and accurate detection. The study proposes an innovative hybrid machine learning approach by combining XGBoost with artificial neural network (ANN) to predict the risk of cardiovascular disease. For this purpose, the methodology employed a large dataset comprising 70,000 patient records having 12 useful features. Alongside, other data preprocessing techniques like mean arterial pressure and body mass index have been employed such as missing value imputation, outlier detection and engineering of additional features. Considering possible gender-specific risk factors, k-modes clustering partitions the dataset according to the given attribute in it. A stratified 70–30 split for training the model with hyperparameter tuning is used in XGBoost and ANN models through ten-fold cross-validation method. Having an impressive accuracy of 85.65%, this hybrid approach outperforms traditional machine learning techniques such as logistic regression, decision trees, support vector machines etc. The other advantage about our model is that it has got high precision (85%), recall (81%) and F1-score (75%). XGBoost-ANN combination further provides interpretability benefits which explain key risk factors driving personalized treatment plans for patients much better than any other alternative predictive model does. This work shows how well a proposed hybrid machine learning approach can accurately predict cardiac diseases improving diagnostic capabilities in healthcare decision-making process.