Cardiovascular diseases (CD), or heart diseases (HD), lead to approximately 17.9 million deaths each year, constituting 32% of global fatalities. Early detection and appropriate treatment of HDs can significantly reduce mortality rates, with timely intervention before disease progression enhancing treatment efficacy. Early detection is achievable through routine medical examinations and monitoring key symptoms, such as cholesterol levels, blood pressure variations, diabetes and obesity. This manuscript introduces a heart disease prediction (HDP) model designed to identify the presence of HDs at an initial stage. The study explores three methodologies: (a) traditional convolutional neural network (CNN), (b) CNN augmented with flower pollination optimization (FPO) algorithm and (c) CNN combining Teaching Learning-Based Optimization (TLBO) coupled with genetic algorithm (GA) for refined HDP. The model progresses through stages of data preparation, model construction, training and evaluation. The traditional CNN model resulted an accuracy of 81.97%, precision of 84%, recall of 81% and F1-score of 83%. Incorporating the FPO algorithm, model's performance is enhanced with accuracy, precision, recall, F1-score of 85.25%, 90%, 81% and 85%, respectively. Further, optimized with TLBO and GA hybrid approach lead to superior performance with an accuracy, precision, recall, F1-score of 86.9%, 87.5%, 87.5% and 87.5%, respectively. The area under curve (AUC) for the receiver operating characteristics (ROC) and precision–recall curve (PRC) highlights the performance of the proposed hybrid methodology. These outcomes underscore the effectiveness of merging bio-inspired algorithms with CNN for early stage HD prediction (HDP), offering significant advancements in healthcare diagnostics.

Enhanced heart disease prediction through hybrid CNN-TLBO-GA optimization: a comparative study with conventional CNN and optimized CNN using FPO algorithm