مشخصات پژوهش

Cardiovascular diseases are among the leading causes of mortality worldwide. These diseases are often caused by complex factors such as unhealthy lifestyles, heredity, and metabolic disorders. Early and accurate diagnosis of these diseases can help reduce mortality rates and improve patients' quality of life. In this context, the use of intelligent technologies and advanced machine learning methods in medical data analysis has gained significant importance. With the widespread use of medical data, the volume and complexity of this data have dramatically increased [1, 2]. This growth has caused traditional machine learning algorithms to face issues such as computational complexity, high processing time, and limited performance in noisy data conditions. To improve the performance of machine learning models, efficient dimensionality reduction becomes essential [3]. In this study, supervised autoencoders (SAEs) are employed as a novel and effective method for dimensionality reduction and feature extraction. Unlike traditional autoencoders that focus solely on reconstructing input data, SAEs incorporate label information (Labels) and reconstruction errors to learn more distinctive and optimized latent representations [4]. As explained by Lei et al [5]., this approach enables the creation of latent representations that are not only effective in reconstructing data but are also highly beneficial for classification. This significantly enhances the performance of machine learning models, especially when dealing with noisy and complex datasets. In addition to supervised autoencoders, Principal Component Analysis (PCA) is also utilized as a classical method for dimensionality reduction and performance comparison. PCA is one of the most common dimensionality reduction techniques, mapping data to a lower-dimensional space by identifying its principal components. It is often used as a benchmark to evaluate the effectiveness of new methods. While PCA is effective in extracting lin