<![CDATA[Heart disease is the leading cause of death worldwide. To detect the risk of heart disease early, an accurate and efficient classification method is needed. This study proposes a hybrid approach by combining Information Gain (IG) feature selection and the Backpropagation Neural Network (BPNN) classification algorithm. The dataset used is the Heart Disease Dataset from the UCI Repository, consisting of 303 patient records. Eight top features were selected using Information Gain. The BPNN model was trained using parameters hidden_layer_sizes=(16, 8), activation='relu', and learning_rate_init=0.01. A hidden layer with 16 and 8 neurons enables the network to learn complex patterns in the data. The ReLU (Rectified Linear Unit) activation function is used to speed up training convergence and avoid the vanishing gradient problem. The learning_rate_init=0.01 parameter controls the speed of weight updates during the learning process, affecting the model's stability and convergence. The evaluation results show that the model achieves 79.12% accuracy, 84.44% precision, 76.00% recall, 80.00% F1-Score, and 85.95% AUC. The 5-Fold Cross Validation yielded an average accuracy of 82.15%. These results indicate that the IG + BPNN hybrid approach provides good and stable classification performance in detecting heart disease. Keywords: Heart Disease, Information Gain, Backpropagation Neural Network, Classification, Data Mining]]>
Copyrights © 2025
