An automatic heart disease prediction using cluster-based bi-directional LSTM (C-BiLSTM) algorithm

Heart disease involves many diseases like block blood vessels, heart attack, chest pain or stroke. Heart disease will affect the muscles, valves or heart rate, and bypass surgery or coronary artery surgery will be used to treat these problems. In this paper, UCI heart disease dataset and real time dataset are used to test the deep learning techniques which are compared with the traditional methods. To improve the accuracy of the traditional methods, cluster-based bi-directional long-short term memory (C-BiLSTM) has been proposed. The UCI and real time heart disease dataset are used for experimental results, and both the datasets are used as inputs through the K-Means clustering algorithm for the removal of duplicate data, and then, the heart disease has been predicted using C-BiLSTM approach. The conventional classifier methods such as Regression Tree, SVM, Logistic Regression, KNN, Gated Recurrent Unit and Ensemble are compared with C-BiLSTM, and the efficiency of the system is demonstrated in terms of accuracy, sensitivity and F1 score. The results show that the C-BiLSTM proves to be the best with 94.78% accuracy of UCI dataset and 92.84% of real time dataset compared to the six conventional methods for providing better prediction of heart disease.