Crowding Reduction And Waiting Time Analysis In Health-Care System Using Machine Learning
| dc.contributor.advisor | Olawoyin, Richard | |
| dc.contributor.author | Hijry, Hassan Mohmmed | |
| dc.contributor.other | McDonald, Gary | |
| dc.contributor.other | Edward, William | |
| dc.contributor.other | Debnath, Debatosh | |
| dc.date.accessioned | 2022-11-15T17:54:33Z | |
| dc.date.available | 2022-11-15T17:54:33Z | |
| dc.date.issued | 2022-05-20 | |
| dc.description.abstract | In the hospital setting, the emergency room (ER) offers timely emergency care for patients and is considered the busiest department because of the urgency of cases. Emergency rooms have the highest number of patients overcrowding within any hospital; more than 50% of the patients admitted to the hospital come through the ER. Healthcare management is continuously trying to minimize wait times and optimize the hospital's allocated resources, but most ERs still suffer from the overcrowding crisis due to the stochastic arrival and random arrival distribution. Advanced techniques, such as machine learning algorithms, are useful for determining real life queue scenarios and patient flow (e.g., waiting time in queue and length of stay), which are considered measures of ER overcrowding. As such, we began by building a model to predict patient length of stay through predictive input factors such as patient age, mode of arrival, and patientÕs type of condition using three machine learning algorithms (e.g., artificial neural networks (ANN), linear regression, and logistic regression). The best model accuracy ANN resulted in an increase of 19.5% compared to the performance from previous studies. Then, the Deep Learning Model was applied for historical queueing variables to vi predict patient waiting time in a system alongside, or in place of, queueing theory (QT). Four optimization algorithms (SGD, Adam, RMSprop, and AdaGrade) were applied and compared to find the best model with the lowest mean absolute error. The results showed that the SGD algorithm achieved better prediction accuracy than the traditional approach and reduced the use of assumptions. Moreover, the model decreased the error reduction by 24% when compared to prior literature. Lastly, we proposed a model to predict the patient waiting time based on the lab test results. Multi-algorithms were implemented by using real-life COVID-19 test results data recorded during the pandemic. Among the eight proposed models, the results showed that decision tree regression performed better for predicting waiting times. Based on experiments performed in the research, this dissertation provides a guideline for waiting time analysis in the queue--not only in healthcare, but also in other sectors, considering model understandability and the feature extraction process. | |
| dc.identifier.uri | http://hdl.handle.net/10323/12012 | |
| dc.relation.department | Engineering | |
| dc.subject | Industrial engineering | |
| dc.subject | Crowding problem | |
| dc.subject | Emergency department | |
| dc.subject | Length of stay | |
| dc.subject | Machine learning | |
| dc.subject | Patient lab test | |
| dc.subject | Patient waiting prediction | |
| dc.title | Crowding Reduction And Waiting Time Analysis In Health-Care System Using Machine Learning | |
| dc.type | Dissertation |