Deep Stacked Ensemble Learning Model for COVID-19 Classification

COVID-19 is a growing problem worldwide with high mortality rate. As result, the World Health Organization (WHO) declared it pandemic. In order to limit spread of disease, fast and accurate diagnosis required. A reverse transcript polymerase chain reaction (RT-PCR) test often used detect disease. However, since this time-consuming, chest computed tomography (CT) or plain X-ray (CXR) sometimes indicated. The value automated that saves time money by minimizing human effort. Three significant contributions are made our research. Its initial purpose use essential finetuning methodology action efficiency variety vision models, ranging from Inception Neural Architecture Search (NAS) networks. Second, plotting class activation maps (CAMs) for individual networks assessing classification AUC-ROC curves, behavior these models visually analyzed. Finally, stacked ensembles techniques were provide greater generalization combining finetuned six ensemble neural Using ensembles, improved. Furthermore, model created all obtained state-of-the-art accuracy detection score 99.17%. precision recall rates 99.99% 89.79%, respectively, highlighting robustness ensembles. proposed approach performed well in lesions on CXR according experimental results.