A Combinatorial Approach: Datamining and an Efficient Deep Neural Network for Heart Disease Prediction

Abstract

This paper explores the role of Data mining and Artificial Intelligence in the area of medical research. Prevention of heart disease is one of the vital areas in medical research. The objective of this study is to design a diagnostic prediction system that can detect and predict heart disease at an early stage by mining relevant data from the clinical dataset using datamining, statistics, and deep learning techniques. Data preprocessing is performed in multiphase such as removing of missing data, numeric transformation, and data normalization for mining an efficient data. Our main contribution is to design an efficient deep neural network model for the early prevention of heart disease. In this point of view, we have designed heart disease prediction system consists of two deep learning neural network architectures namely, (i) Deep neural network in Recognition of heart disease (DeepR) and (ii) efficient Deep neural network in Recognition of heart disease (eDeepR). In these two proposed architectures, DeepR generates 97.64% accuracy and eDeepR generates 99.53% accuracy for recognition of heart disease, after recognition which can be applicable for prevention. To evaluate the performance of proposed networks, conducted experiments on the Cleveland heart disease data set from the UCI repository. Results of proposed systems demonstrate the performance is superior to the previously reported prediction techniques.