[This article belongs to Volume - 56, Issue - 01, 2024]
Gongcheng Kexue Yu Jishu/Advanced Engineering Science
Journal ID : AES-27-05-2024-52

Title : AN EFFICIENT DEEP LEARNING-BASED LSTM MODEL FOR CLASSIFICATION AND PREDICTION OF HEART DISEASE (HD)
Subhash Chandra Jat, Manita

Abstract :

Among the most vital components of existence is healthcare. Heart disease (HD) is a major killer in the modern world. The sickness claims the lives of countless people annually. Worldwide, 17.9 million people succumb to heart disease every year, according to the WHO. The application of image classification, in conjunction with the many HD detecting technologies and methods, might further enhance the outcomes. Thanks to developments in deep learning and machine learning, complex models for automatic early diagnosis of cardiac illnesses have been developed. This is all made feasible by the availability of large-scale data utilised for medical diagnostics. Heart disease identification using a DL method based on image classification is the goal of this research. At the moment, the method of choice for image recognition classification is a deep LSTM. The proposed model is tested using the publicly available Kaggle UCI heart-disease dataset, which includes 14 characteristics and 1050 cases. The SMOTE oversampling approach was used to address the issue of this dataset's extreme imbalance. One drawback of classical methods is their inability to generalise well to data that is different from the training set. Observable differences between training and test accuracies corroborate this. Our model attained a validation accuracy of 97% when several performance measures were utilized to evaluate a proposed algorithm. These metrics included accuracy, precision, recall, and the F1 measure. A proposed architecture is shown successful by a detailed comparison of the model's accuracy with different classification algorithms utilising multiple performance metrics. The work's implementation findings propose that the proposed technique outperforms prior efforts in early HD prediction.