Heart Disease Prediction Using Machine Learning with Greater Cane Rat Algorithm-Based Hyperparameter Tuning

Cardiovascular diseases remain a leading global cause of mortality, highlighting the need for reliable and efficient tools for early risk identification. Machine learning (ML) models have shown promise for heart disease prediction; however, their performance can be highly sensitive to hyperparameter selection. Conventional hyperparameter optimisation approaches (e.g., grid search and random search) may be computationally demanding and can explore the search space inefficiently, especially when hyperparameters interact. To address this, this study investigates the Greater Cane Rat Algorithm (GCRA), a nature-inspired metaheuristic, as a hyperparameter optimisation framework for heart disease prediction. GCRA was applied to tune five classifiers, logistic regression (LR), random forest (RF), support vector machine (SVM), k-nearest neighbours (KNN), and decision tree (DT), while naïve Bayes (NB) was evaluated using default settings due to its limited tunable hyperparameters. Experiments were conducted on the UCI Heart Disease dataset under two train–test splits (80/20 and 70/30). Model performance was assessed using accuracy, precision, recall, and F1-score, and results were compared against genetic algorithm (GA) tuning as well as grid search (GS) and random search (RS). Across most models, GCRA improved performance relative to untuned baselines, with the largest gain observed for KNN, where accuracy increased from 0.8033 to 0.8689 under the 80/20 split. Overall, the findings indicate that GCRA can provide an effective and practical hyperparameter optimisation strategy for ML-based heart disease prediction under consistent evaluation settings.