Applying Machine Learning and Ensemble Learning in Research of Predicting Orthopedic Surgery Duration

Abstract

This study analyzes the prediction accuracy and application contexts of different machine learning models for predicting surgery duration in the orthopedic operating room of CY Hospital. The prediction accuracy of nursing staff for surgery duration was only 22.8%, with an error underestimation rate as high as 66.5%, indicating a substantial bias in the prediction process. The study consists of two parts: the first part involves training and evaluating the performance of individual models, including Multilayer Perceptron (MLP), K-Nearest Neighbors (KNN), and Decision Tree (DT), with results showing that MLP provides the highest accuracy and stability. The second part applies the Voting algorithm to further enhance predictive performance, achieving a final accuracy rate of 95.5%, which is approximately 72.7% higher than that of the nursing staff. The results demonstrate that the Voting algorithm effectively integrates the strengths of multiple base models, maintaining stable predictive performance across various scenarios. Based on specific requirements, this study recommends adopting the Voting model in applications with high accuracy demands, while in resource-limited situations, KNN or DT models can be selected to balance efficiency and accuracy.

Ähnliche Arbeiten