Early diagnosis of cardiac disease using Xgboost ensemble voting-based feature selection, based lightweight recurrent neural network approach

Abstract

Cardiovascular disease (CVD) causes the heart and blood vessels to fail, often resulting in death or stroke. Therefore, early automatic identification of CVD can rescue many lives. CVD identification and prognosis are essential clinical tasks to ensure precise classification results, which assist cardiologists in providing suitable patient treatment. The use of Deep Learning (DL) in the medical field is increasing as it can determine patterns in data. Despite that, CVD prediction is a profound challenge in clinical data analysis. Conventional methods cannot handle hidden patterns, leading to less accurate model predictions. There is a critical need for a new technique that can rapidly and reliably predict future outcomes in patients with CVD. To combat this issue, this research uses a benchmark dataset to present a Lightweight Recurrent Neural Network with a Long Short Term Memory (LRNN-LSTM) method for CVD. Initially, the Min-Max Batch Normalization (M2BN) method is used to verify the ideal margin of collected data values in the dataset. Secondly, they employed the Decision Tree (DT) technique to select the best gain attribute for predicting CVD. Furthermore, the XGBoost Ensemble Voting-based Feature Selection (XGB-EVFS) method determines the profound features of CVD. Then, our proposed LRNN-LSTM algorithm is used to categorize the CVD result to reduce misdiagnosis. The proposed system will develop a model that can accurately predict CVD to decrease mortality from cardiac disease. Therefore, the experiment analysis produces high classification accuracy, precision, and recall with fewer false scores than traditional methods.