Article Preview
Top1. Introduction
Heart disease is the most prominent disease, which affects many people during their old or middle age, and in several cases, it ultimately causes deadly difficulty (Shah, et al., 2020; Fitriyani, et al., 2020). Congestive heart malfunction is heart failure, and it occurs while the heart does not pump enough blood for body needs. Further, heart disease includes the threat factors like obesity, high blood pressure, sleep apnea, alcohol addiction, being inactive, vitamin deficit, heavy metal toxicity, heart attack, smoking, and an improper diet (i.e.), together with animal fats and salt. Medical specialists can identify the damages in the patient's heart and discovers information about the blood pumps in the patient's heart. Furthermore, heart diseases are common in men than women (Ali, et al., 2019). Based on the WHO statistics, non-communicable illnesses, such as heart disease, are responsible for 24% of mortality in India (Pan, et al., 2020). Heart disease is responsible for one-third of mortality worldwide. In other developed countries like the United States, half of the deaths are occurred due to heart ailments. Every year, around 17 million people die from CVD and the disease is particularly widespread in Asia. Furthermore, the CHDD has measured the defacto database for heart disease research (Khedr, et al., 2021).
Wearable sensors and medical testing are used to detect CVD. Nonetheless, extracting important heart disease risk indicators from electronic medical testing is difficult since the doctors attempt to identify patients properly and fast. Furthermore, due to regular medical testing, these EMRs are continually growing in size and becoming more unstructured (Sarmah, 2020; Khan, 2020). At present, wearable sensors are also used for monitoring the patient’s body, both externally and internally to identify heart disease. Nevertheless, the wearable sensor data are degraded by signal artifacts like missing values and noise that minimizes the system performance and produces inaccurate outcomes for heart disease prediction. Using EMRs and wearable sensors together is a crucial and challenging task in the first place while monitoring cardiac patients (Shankar, et al., 2020; Rani, et al., 2021). Second, collecting meaningful and relevant characteristics data is a challenge for detecting heart disease (Prakash, et al., 2019). As a consequence, the intelligent systems automatically integrate information gathered from both EMRs and sensor data, and it could evaluate the retrieved data to discover hidden signs of problems in heart and forecast heart illness (Nandy, et al., 2021).
Certain variables of risk factors make the CHD prediction more complex, and it maximizes the treatment and diagnosis cost (Dwivedi, 2018; Ali & Bukhari, 2021). For resolving the diagnosis cost and its complexities, advanced ML models are widely used by researchers for predicting the CHD from the patient’s clinical data. The prediction performances of more ML models (Latha, et al., 2019; Shaik & Ganesh, 2020; Chithra & Jagatheeswari, 2019) like RBF and MLP are evaluated for predicting the CHD. The most efficient method is MLP that yields an area in the ROC curve (Samuel, et al., 2020; Baggen, et al., 2019). Furthermore, the hybrid forward selection approach is used since it picks minor subgroups and improves CVD incidence accuracy with fewer features. Certain groups reported that approaches like FL, DL, and ANN methods (Cristin, et al., 2019; Xu, 2020; Bhagyalakshmi, et al., 2018) are more helpful to enhance the heart disease diagnosis. Moreover, the data mining approach is utilized in the healthcare industry as it takes less time for the prediction of disease with more accurate outcomes (Chu, et al., 2020; Adabag & Langsetmo 2020). NN (Dattatraya & Raghava Rao, 2020; Gupta, 2020) is normally regarded as the most excellent tool for the prediction of diseases such as brain disease and heart disease (Rajakumar, 2013; Rajakumar, 2013; Swamy, et al., 2013; Rajakumar & George 2013; Rajakumar & George, 2012). The following are the key contributions of the accepted model: