Optimizing Overnight Patient Care: An Investigation into the Integration of Hybrid Machine Learning Algorithms

Abstract

In the medical field, it is determined if a patient needs to stay overnight (in-care) or not (out-care) after a medical operation or surgery by analyzing their diagnostic information. During this procedure, the medical personnel evaluate the patient's vital signs, such as age, gender, hemoglobin levels, and red blood cell count, to determine whether the patient needs to be kept under overnight monitoring in the hospital or whether they can be discharged to go home. Assessing a model's effectiveness in data science applications in the medical industry, precision is considered a superior criterion. The reason for this is that, in the context of human health, it is crucial to have an algorithm that has a low False Negative Rate. By evaluating Precision (rather than Accuracy), we guarantee consistency. While the aforementioned methods provide satisfactory accuracy, they do not evaluate performance based on precision and recall. In this research, we introduced a novel phase of a probability-based decision support system to compare the efficacy of different classification models. The improved DSS yields superior outcomes due to the implementation of a diverse range of probability thresholds. This enhances precision and recall, resulting in an overall improvement in the model's performance with minimal impact on accuracy