An Efficient Early Stage Heart Disease Risk Detection Using Machine Learning Techniques

In the medical field, early prediction of disease is a big challenge. This paper focuses on predicting heart disease at an early stage. Heart disease is a fatal human disease that rapidly increases at a global level. This disease affects both developed as well as undeveloped countries which subsequently causes death. In heart disease, the heart doesn't supply the required volume of blood to other body parts. It is essential to diagnose this disease at the early stage for preventing patients from higher damage. In medical diagnostic systems, errors can cause improper medical treatments which can result in the death of the patient. Artificial Intelligence (AI) can be applied in several healthcare processes to minimize the time and resources required in examining and diagnosing patients. In AI, machine learning has upsurged as an important technique in diagnosing heart disease. This paper showcases the current state-of-the-art techniques utilized in heart disease prediction. This paper proposes an architecture for heart disease prediction by using machine learning techniques along with Principal Component Analysis (PCA) for dimensionality reduction. It utilizes a standard UCI dataset of Kaggle having a rich set of attributes. Several standard machine learning techniques are utilized in the proposed architecture. The paper showcases the comparison of different machine learning algorithms for the detection of heart disease using standard parameters such as classification accuracy, precision, recall, an area under curve (AUC), F1 measure and ROC curve. It depicts that the Naive Bayes classifier outperforms for training without feature reduction and with feature reduction. However, Adaboost outperforms in testing in the proposed architecture.