Building an Efficient Heart Disease Prediction System by using Clustering Techniques

Abstract

Listen

Translate article

The healthcare management system (HMS) is a key to successful management of healthcare facilities such as clinics or hospitals. In today’s world, heart disease is one of the most common causes of illness and death. Heart disease proves to be the leading reason for the death of a person. Heart disease affects human life very badly and its prediction is a complex task. Heart disease predicts a system which can assist medical professionals in predicting the status of a patient’s heart condition based on their clinical data or records. We choose 14 attributes such as age, gender, cholesterol, type of chest pain, fasting blood sugar, resting blood pressure (restbps), resting electro cardio graph (ECG (for heart monitoring)), exercise-induced angina, maximum heart rate, slope, old peak and number of vessels coloured. In this paper, we proposed a work to predict the likelihood of heart disease and classify patient risk level using various hybrid techniques. We combined unsupervised and supervised approaches to achieve a higher-level accuracy of classification. The unsupervised approach is very important in the hybrid learning methods. Hence, K-means clustering used for patient segmentation is helpful for grouping of similar types of patients within a cluster. The holdout approach is used to form the training and testing data from each cluster. Then, the model is generated by various classification algorithms such as DT, NB, SVM and KNN. The heart disease predicts at early stage based on risk factors, and this experiment demonstrated that combining the K-means algorithm along with a decision tree improves accuracy.