摘要
基于上海某医院一年的手术时长数据，采用三相位的超伽马（Hypergamma）概率分布模型拟合所有手术、各科室、各种类三种分类层级手术时长的经验概率分布。通过最大期望算法（Expectation Maximization，EM）和遗传算法（GeneticAlgorithm，GA）对拟合模型进行参数估计。以均方根误差（RMsE）作为拟合效果的衡量指标，数值实验结果表明超伽马概率分布在三种分类层级手术时长分布的拟合上，性能优于对数正态（Lognormal）、伽马（Gamma）和威布尔（Weibull）三种经典分布。在超伽马概率分布模型的参数估计上，设计的GA算法在求解效率上优于经典的EM算法。
Based on one-year surgery duration data of ahospital in Shanghai, a Hypergamma probability distribution model was applied in the fitting of the empirical surgery duration distributions of all surgeries, surgeries from various departments, surgeries of different kinds. The parameters of the fitting model were estimated by Expectation Maximization（EM）algorithm and Genetic Algorithm（GA）respectively. The root mean square error（RMSE）wasapplied to measure the fitting accuracy,and numerical results showed that the Hypergamma probability distribution model was more suitablein the fitting of the surgery duration distributions than the classical probability distributions such as Lognormal, Gamma and Weibull Distributions. The proposed GA was superior to the classical EM algorithm in the parameters＇ estimation efficiency.
出处
《工业工程与管理》
CSSCI
北大核心
2018年第1期51-58,共8页
Industrial Engineering and Management
基金
中华人民共和国科学技术部创新方法工作专项资助项目(2015IM030200)
关键词
Hypergamma概率分布
手术时长
最大期望算法
遗传算法
hypergamma
probability distribution
surgery duration
expectationmaximization algorithm
genetic algorithm
