Rand Similarity Feature Matching Based Silhouette Baum–Welch Mean Shift Data Clustering for Disease Prediction with Big Data

Abstract

Disease prediction from huge medical data is useful to predict the status of health result to the patient. Predicting the risk causes a possible prediction of disease and creates a preventive measure for early-stage disease diagnosis and limits the mortality. Many previous works have been introduced for disease prediction. However, better accuracy and time minimization are the challenging issues in big data analysis. To enhance the disease prediction accuracy, a novel method called a Rand Similarity Feature Matching based Silhouette Baum–Welch Meanshift Data Clustering (RSFM-SBWMDC) is introduced. The main of the RSFM-SBWMDC method is to enhance accuracy and reduce time consumption. RSFM-SBWMDC method consists of three processes namely feature selection, clustering, and cluster validation. At first, Rand Similarity Feature Matching is applied for selecting the significant features. The similarity matching score between the features is high, and then the feature is selected to predict the disease. Then, Baum–Welch Mean Shift Data Clustering process is carried out to group the patient data with the selected features based on the maximum likelihood estimation. After clustering the patient data, the silhouette index score is calculated for every clustered patient data to measure how similar the data is to its cluster. This helps to minimize the irrelevant data grouped and improving the accurate prediction. The results analysis shows that the RSFM-SBWMDC method provides better performance as compared to conventional to works.