A Data-Driven Heart Disease Prediction Model Through K-Means Clustering-Based Anomaly Detection

Heart disease, alternatively known as cardiovascular disease, is the primary basis of death worldwide over the past few decades. To make an early diagnosis, a data-driven prediction model considering the associate risk factors in heart disease can play a significant role in healthcare domain. However, to build such an effective model based on machine learning techniques, the quality of the data, e.g., data without “anomalies” or outliers, is important. This research investigates anomaly detection in the healthcare domain to effectively predict heart disease using unsupervised K-means clustering algorithm. Our proposed model first determines an optimal value of K using the Silhouette method to form the clusters for finding the anomalies. After that, we eliminate the identified anomalies from the data and employ the five most popular machine learning classification techniques, such as K-nearest neighbor, random forest, support vector machine, naive Bayes, and logistic regression to build the resultant prediction model. The efficacy of the proposed methodology is justified using a standard heart disease dataset. We also take into account the data plotting to test the exactness of the detection of anomalies in our experimental analysis.