Identification of Significant Clinical Attributes for Developing Heart Disease Prediction System

Abstract

The doctor uses a variety of laboratory testing, physical exams, and occasionally even invasive tests to identify diseases. To identify the disease in its early stages, a physician with exceptional training, experience, and domain expertise is required. Medical professionals may benefit from using diagnostic tools based on machine learning. Medical tests, both invasive and non-invasive, are required for the diagnosis of heart disorders. For the Indian population, prediction models for heart disease based on non-invasive clinical features will be very helpful. Affordable, accessible, and high-quality healthcare is still out of reach for a large portion of the population in India. The lack of infrastructure in rural locations makes it difficult to provide early disease diagnosis and treatment, which delays care, increases morbidity, and increases mortality. In India during the past two decades, the mortality rate from non-communicable diseases has increased alarmingly. These forecasting models were developed using four distinct machine learning procedures: logistic regression, k-NN, Support vector machine, and Random Forest. Numerous combinations of clinical indicators were thought of. The most crucial features were those that boosted performance in tandem. Important clinical factors were identified in this study to include gender, age, BMI, hypertension, diabetes, alcohol use, smoking, family history, total cholesterol, inactivity, healthy eating habits, stress, and anxiety. A random forest-based system achieved 91.2 percent accuracy, 93.5 percent specificity, and 92.5 percent sensitivity. The primary clinical characteristics utilised in creating a low-cost and easily accessible CVD prediction system. It has been suggested that similar research be conducted on large datasets obtained from other universities. This could help researchers find even more potentially important clinical traits.