Optimasi Model Artificial Neural Network menggunakan Genetic Algorithm untuk Peningkatan Klasifikasi Prediksi Penyakit Jantung

Abstract

Heart disease is one of the leading causes of death, so an accurate prediction system is needed to support early detection. This study optimizes an Artificial Neural Network (ANN) using a Genetic Algorithm (GA) and compares it with an unoptimized ANN and grid search (GS). The dataset used is the Heart Disease dataset (Kaggle), which contains 1,025 samples with 14 features. The data was split 70:30 for training and testing, numerical features were normalized using StandardScaler, and the model built was an Artificial Neural Network model for binary classification. GA optimized the number of neurons in two hidden layers, dropout, and learning rate with a population of 10, a maximum of 20 generations, and early stopping if accuracy did not improve within 5 generations. Evaluation of test data using accuracy, precision, recall, and F1-score. The results show a significant improvement. ANN without optimization achieved an accuracy of 78.57%, and ANN with grid search (GS) achieved an accuracy of 87.66%, while ANN optimized with Genetic Algorithm (GA) achieved 92.86% (precision 92.05%, recall 93.29%, F1 92.67%). The best configuration obtained by the Genetic Algorithm was 62 neurons in the first layer, 14 neurons in the second layer, a dropout rate of 0.303020202, and a learning rate of 0.002163161. These findings confirm that GA effectively explores the hyperparameter space adaptively and improves model generalization on heart disease data.