A surgery duration prediction method based on reinforcement learning hyperparameter optimization

By using a hyperparameter optimization method based on reinforcement learning, and employing TabTransformer and PRO reinforcement learning algorithms to optimize the operation duration prediction model, the problems of low prediction accuracy and poor cross-departmental adaptability in existing technologies are solved, achieving more efficient resource utilization and more stable prediction results.

CN122245614APending Publication Date: 2026-06-19HEFEI UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HEFEI UNIV OF TECH
Filing Date
2026-01-27
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing machine learning models have low accuracy in predicting surgical duration, lack adaptability, and have weak cross-departmental generalization ability, resulting in low resource utilization efficiency and increased costs.

Method used

A reinforcement learning-based hyperparameter optimization method is adopted, which combines the TabTransformer model and the PRO reinforcement learning algorithm with the performance prediction model PMLP to optimize the hyperparameter combination and achieve adaptive and rapid response of cross-departmental data.

Benefits of technology

It improves the accuracy and stability of operation duration prediction, reduces training costs and computing resource requirements, and enhances the efficiency of operating room resource utilization and the overall benefits of the hospital.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122245614A_ABST
    Figure CN122245614A_ABST
Patent Text Reader

Abstract

This invention discloses a surgical duration prediction method based on reinforcement learning hyperparameter optimization. First, a benchmark surgical dataset is acquired for training the master prediction model. Then, multiple statistical features of historical surgical datasets and the historical optimal hyperparameter combinations of the master prediction model are obtained. Next, the similarity between the benchmark and historical surgical datasets is calculated. Based on the comparison between the similarity and a similarity threshold, either the historical optimal hyperparameter combination or the PRO reinforcement learning algorithm is used to obtain initial candidate hyperparameter combinations. Then, the PRO reinforcement learning algorithm combined with a performance prediction model is used to optimize the hyperparameters of the master prediction model to obtain the optimal hyperparameter combination. Finally, the predicted surgical duration data corresponding to the current predicted surgical task data is obtained by loading the master prediction model with the optimal hyperparameter combination. This invention reduces training overhead and significantly improves the efficiency of hyperparameter optimization while maintaining prediction accuracy.
Need to check novelty before this filing date? Find Prior Art