Modeling patient flow in the emergency department using machine learning and simulation

Recently, the combination of machine learning (ML) and simulation is gaining a lot of attention. This paper presents a novel application of ML within the simulation to improve patient flow within an emergency department (ED). An ML model used within a real ED simulation model to quantify the effect of detouring a patient out of the ED on the length of stay (LOS) and door-to-doctor time (DTDT) as a response to the prediction of patient admission to the hospital from the ED. The ML model trained using a set of six features including the patient age, arrival day, arrival hour of the day, and the triage level. The prediction model used a decision tree (DT) model, which is trained using historical data achieves a 75% accuracy. The set of rules extracted from the DT are coded within the simulation model. Given a certain probability of free inpatient beds, the predicted admitted patient is then pulled out from the ED to inpatient units to alleviate the crowding within the ED. The used policy combined with adding specific ED resources achieve 9.39% and 8.18% reduction in LOS and DTDT, respectively.

Performance Prediction for Convolutional Neural Networks in Edge Devices

Running Convolutional Neural Network (CNN) based applications on edge devices near the source of data can meet the latency and privacy challenges. However due to their reduced computing resources and their energy constraints, these edge devices can hardly satisfy CNN needs in processing and data storage. For these platforms, choosing the CNN with the best trade-off between accuracy and execution time while respecting Hardware constraints is crucial. In this paper, we present and compare five (5) of the widely used Machine Learning based methods for execution time prediction of CNNs on two (2) edge GPU platforms. For these 5 methods, we also explore the time needed for their training and tuning their corresponding hyperparameters. Finally, we compare times to run the prediction models on different platforms. The utilization of these methods will highly facilitate design space exploration by providing quickly the best CNN on a target edge GPU. Experimental results show that eXtreme Gradient Boosting (XGBoost) provides a less than 14.73% average prediction error even for unexplored and unseen CNN models' architectures. Random Forest (RF) depicts comparable accuracy but needs more effort and time to be trained. The other 3 approaches (OLS, MLP and SVR) are less accurate for CNN performances estimation.