Article Preview
TopIntroduction
In the fourth industrial revolution, technology in the health sciences is rapidly evolving. Every day, the population grows, and new disease outbreaks are becoming more consistent. If one needs proof, COVID-19 is the best example of a new sickness that caused the entire planet to shut down state due to the excruciating condition. Technological advancements imply not only the creation of new software but also the upgrading of existing software and it entails enhancements to data analysis (Anderson & Tushman, 1990). Every day, new patients arrive, and in the mind of a newborn kid, they require the care of a neonatal intensive care unit (NICU) in order to overcome unusual disorders or make fundamental changes to the disease with which they were born (Wigert at el., 2006).
Patient data are crucial for determining the consequences of new technologies or identifying disease patterns (Gregory & Radovinsky, 2012). Prior to the different thresholds, each patient's data may vary, However, if an alike sickness, the existing dataset can be compared with the new data. If the ailment is coronary heart disease, any doctor may easily recognize hypertension as one of the essential characteristics that might lead to heart disease in any person. Multiple persons with the same sickness won't have the same symptoms. In this contrast, one of the most extensively used strategies for detecting disease patterns is collecting data and observing patterns in several cases (Ofran at el., 2012). Health information or data management is a key methodology that is widely adopted for simulation or finding the causes or finding the solution to any medical treatment.
Machine learning is nowadays widely employed in medical science and supervised machine learning algorithms are primarily used to uncover patient sickness patterns and sort out the key features for which a disease occurs most frequently (Kolachalama & Garg, 2018). Machine learning supervised algorithms named Logistic regression, Support vector machine, Decision tree, Random forests, Naive Bayes is commonly employed to assist clinicians in data analysis (Beunza at el., 2019). Machine learning algorithms benefit the health professional by taking into account the dataset variables to observe multiple patient symptoms and disease factors. There may be some strange scenarios when the patient's data does not match that of other patients, but in the vast majority of cases, it does. In several circumstances, it has been demonstrated that there can be exceptions in the patient data. Machine learning models can easily predict or evaluate the risk of a patient’s sickness reasons. It can help to identify the future conditions of a patient by adopting the patient’s previous history or data.
Deep learning is rapidly implemented in medical sciences, due to its multiple hidden layer's evaluations and accuracy in prediction. Image classifications or identification of MRI, X-RAY, Ultrasound images, etc., or predicting disease from the images, are vast uses in deep learning nowadays. It may be used for mapping and prediction by training a variety of datasets. Deep learning's popular algorithms that are used in medical sciences for visualization and prediction are Artificial Neural Networks (ANN), Convolution Neural Network (CNN), Recurrent Neural Network (RNN), etc. (Sajeev at el., 2019). In deep learning, the epochs help to find out the best output, which is used to train the dataset over and over again to obtain the best outcome.
In this research, the Framingham CHD dataset was utilized to determine whether a patient has future coronary heart disease risk by creating a model in Machine Learning and Deep Learning. Simultaneously, the effectiveness of Machine Learning and Deep Learning predictions was investigated. The paper aims to visualize the best effective model for predicting the risk of coronary heart disease during a ten-year period through the comparison and visualization of different models' prediction accuracy, AUC score, and ROC curve. For building the model, 70% of the total data was used for train purposes and 30% of the data for testing purposes where the label(y) assigned to the variable of the Framingham datasets indicated Ten Year CHD results. Health data is vastly available nowadays and health professionals are more likely to investigate a disease's reason or solve a disease using different techniques or implementing data in multiple models to predict or identify the reason.