Meta-modeling for cardiovascular risk prediction

The predictive capability of machine learning models for cardiovascular disease risk assessment increases by the integration of Heterogeneous medical datasets. The proposed study harnesses the power of two random forest classifiers trained on two different datasets namely UCI-Combined Heart Disease and Framingham Heart introducing a logistic regression meta which overcomes the generalizability issue posed by the limitations of single dataset models The introduced stacked ensemble framework merges the outputs from both base models through logistic regression. Proposed meta-model achieves an accuracy of 82% which clearly supports superior robustness, handling missing-values and interpretability compared with conventional ensemble methods. The detailed analysis of error correlations favors the framework&s;s improved cross-dataset generalizability and clinical transparency. Results clearly indicates that the cross-dataset ensemble learning can provide a more reliable and adaptable approach for cardiovascular risk prediction.