Heart disease is a major cause of morbidity and mortality worldwide. Early diagnosis and treatment are crucial to prevent serious complications. In recent years, machine learning techniques have been used for heart disease prediction with promising results.
In this study, we propose an automated diagnostic system for heart disease prediction based on a statistical model and optimally configured deep neural network. The proposed system consists of two stages: the first stage uses a logistic regression model to predict the probability of heart disease based on clinical features such as age, sex, blood pressure, cholesterol levels, etc. The second stage uses a deep neural network (DNN) to further improve the accuracy of prediction by learning from raw data.
To optimize the DNN architecture, we use a Bayesian optimization approach that searches for the best hyperparameters (e.g., number of layers, neurons per layer, activation functions) in an efficient manner. We evaluate our proposed system using the Cleveland Heart Disease dataset and compare its performance with several baseline models.
Our experimental results show that our proposed system achieves superior performance compared to other models. Specifically, it achieves an accuracy of 88.35% and an AUC score of 0.93, which are significantly better than those achieved by logistic regression (accuracy: 83.33%, AUC: 0.87) and other baseline models.
In conclusion, our proposed automated diagnostic system can accurately predict heart disease risk and provide clinicians with valuable information for early intervention and prevention strategies. Future work will focus on validating our system on larger datasets and integrating it into clinical practice.
