A Modified Genetic Algorithm–Based Feature Optimization Framework for Cardiovascular Disease Risk Prediction

Abstract

The proper diagnosis of cardiovascular disease (CVD) risk rises as a significant issue in clinical practice because the risk is directly associated with patient survival. Thus, correct and timely assessment of the risk is crucial. This work suggests a CVD probability prediction model, which combines ensemble learning and Deep Learning (DL) algorithms with a feature choice approach supported by a Modified Genetic Algorithm (MGA). A balanced clinical dataset consisting of 1,025 patient records and 14 medically relevant attributes was used which was retrieved via a publicly available repository with benchmarks dataset. The comparison of a broad range of predictive models, including classical Machine Learning (ML) algorithms like Logistic Regression, Support Vector Machines (SVM), K-Nearest Neighbors, Naïve Bayes, Decision Trees, Random Forests, XGBoost, and LightGBM, as well as DL models, including Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Multi-layer Perceptrons (MLP), was made. The MGA based feature collection is used to select the most informative clinical attributes, decrease feature redundancy, and increase discriminative ability. The outcomes of experimental findings prove that the models with the assistance of MGA are most effective as they are more predictive and stable across various classifiers. The results demonstrate the imperative need of intelligent feature selection to enhance model generalization and predictability, which reinforces predictive modelling framework with feature optimization against cardiovascular risk prediction. The paper presents the framework that mediates between optimization methods and predictive modeling to provide valuable information on the next generation of data-driven cardiovascular diagnostics.