Yapay sinir ağı kullanarak COVID-19 hastalarının yoğun bakım ünitesi gerekliliklerinin tahmini

Since the first pneumonia case caused by the new 2019 coronavirus (COVID -19) was found in Wuhan, the impact and consequences of this pandemic, which has affected the whole world, are still ongoing. The healthcare sector is one of the most affected sectors by this pandemic. During the most difficult times of the COVID 19 pandemic, the intensive care unit (ICU) was usually the bottleneck of a medical facility. ICU admission has evolved from a medical decision to a resource allocation issue. One of the great difficulties experienced by health care was the increase in hospital admissions due to the spread of the disease. The need for decision support arises at all levels of health care. The ability to predict what type of individual needs will occur with a positive test, or even earlier, is beneficial for authorities to plan resources effectively. In this study, a decision support framework is proposed for predicting the demand of COVID-19 patients in ICU within the framework of prescriptive analysis and developing a classification model using artificial neural networks. The model has been trained, validated, and tested using a total dataset of 120026 COVID 19 cases. During the test phase, performance is observed using confusion matrices, training and validation loss curves, and other performance metrics such as accuracy, precision, recall, and F1 score. The Synthetic Minority Oversampling Technique (SMOTE) is applied to unbalanced datasets to improve performance. It was predicted with 79% accuracy whether a person requires the Intensive Care Unit.