Introduction
Effective triage in the emergency department (ED) is essential for optimizing resource allocation, improving efficiency, and enhancing patient outcomes. Conventional systems rely heavily on clinical judgment and standardized guidelines, which may be insufficient under growing patient volumes and increasingly complex presentations.
Methods
We developed a machine learning triage model, MIGWO-XGBOOST, which incorporates a Multi-strategy Improved Gray Wolf Optimization (MIGWO) algorithm for parameter tuning. Missing data were processed, and the dataset was randomly split into 80 percent for training and 20 percent for testing. Model performance was evaluated against standard XGBOOST, GWO XGBOOST, AdaBoost, LSTM, and CNN-BiGRU.
Results
MIGWO-XGBOOST improved accuracy by 8.5 percent over unoptimized XGBOOST and reduced optimization time by 9,285 seconds relative to GWO-XGBOOST. Compared with other benchmarks, accuracy gains were 12.5 percent over AdaBoost, 3.3 percent over LSTM, and 1.9 percent over CNN-BiGRU. These results demonstrate both predictive strength and computational efficiency in complex data environments.
Discussion
MIGWO-XGBOOST provides a robust framework for rapid and precise triage decisions in the ED. By enhancing accuracy while substantially reducing computational time, this approach demonstrates the potential of advanced machine learning to support emergency decision-making and optimize patient care pathways.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号