Comparison of the Effectiveness of Different Optimizers with Deep Learning Algorithms for the Prediction of Heart Disease

Heart disease continues to rank among the world's top causes of death, highlights the pressing need for precise and effective prediction models. Using the MIT-BIH Arrhythmia dataset, this paper proposes a deep learning-based system for predicting cardiac illness. Raw electrocardiogram (ECG) signals are preprocessed using the Discrete Wavelet Transform (DWT) to eliminate noise and maintain important signal components. Important statistical features especially the mean and standard deviation are extracted to describe crucial signal characteristics. Lasso regression is used for feature selection, keeping just the most relevant attributes to reduce model complexity and narrow the input space. Because a Gated Recurrent Unit (GRU) network can effectively describe the sequential dependencies seen in ECG data, it is used for categorization. The model's performance is further enhanced by experimenting with several training optimizers, such as Adam, RMSprop, and Stochastic Gradient Descent (SGD). Accuracy, precision, recall, and F1-score are among the common metrics used to assess the efficacy of the suggested strategy. According to the results, the model performs well across various evaluation criteria and has a high classification accuracy, making it a promising tool for heart disease diagnosis and early detection.