Accurate Heart Disease Prediction System Using Optimized Data Mining Techniques

Abstract

Heart disease is the frequently found disease in various peoples which would cause more serious and dangerous effects. Various studies have been projected earlier whose major aim is to predict the heart disease more accurately. In our previous research method Fuzzy Rough Set Theory combined with Support Vector Machine (FRS - SVM) is introduced which can ensures the optimal prediction rate by selecting the risk factors accurately which can lead to improved accuracy rate. However FRS-SVM might lack in its performance in case of presence of more missing values in the database. This research method cannot support the large dimensional dataset which needs to be focused well enough for accurate prediction rate. This problem is resolved in this investigation by introducing the framework namely Heart disease prediction using Alpha Rough Set Theory combined with Fuzzy SVM (ARST-FSVM). In this research method, Modified K-Means clustering algorithm is utilized for preprocessing the input dataset which would avoid the noisy data present in the database. Then missing data value in the database is handled using normalization technique where NLLS imputation is applied. And then feature dimensionality reduction is done using Alpha rough set theory (α-RST) approach. From those reduced feature set, optimal feature selection in terms of relevancy is done using Hybrid Bee colony algorithm with Glowworm Swarm Optimization (HBC-GSO) approach. Finally heart disease prediction is done using classifier namely fuzzy based SVM. The overall research method ensures that the proposed research technique leads to ensure it can direct to most favorable and accurate heart disease diagnosis outcome.