Advance Resource Planning in Hospital Emergency Departments Using Machine Learning Techniques

Abstract

Accidents are likely to happen at workplaces which requires employees to rush to the hospitals for emergency treatment. Due to increase in population, treating various medical cases has led to longer waiting times at emergency treatment units (ETUs). The reasons being the ambulance divergence, less staff, and reduced management. An approach to decrease overcrowding at ETU can be the application of modern techniques. Machine learning (ML) is the one which is used to find patients with high illness, therefore developing models that can avoid jams at ETU. In this paper, a new ML technique, light GBM (LGBM), is implemented to increase the predictions rate based on data gathered from hospitals of Northern Ireland. In addition, the proposed model is compared to other ML models such as decision tree and gradient boosted machines (GBM). Test results indicate that LGBM is more efficient with an accuracy of 86.07%. Also, the time taken to produce future predictions by LGBM was 12 milliseconds, whereas decision tree and GBM took 16 milliseconds and 20 milliseconds, respectively.