基于GBOM和改进SVM的DRGs病种系列成本预测  

Study on Method of Disease Series Cost Estimation of DRGs Mode Based on Improved Support Vector Machine

作　　者：徐靖[1] 刘子先[1] 李竹梅[2] 李惠[1,3] 

机构地区：[1]天津大学管理与经济学部,天津300072 [2]山东省邹城市人民医院,邹城273500 [3]天津中医药大学,天津300193

出　　处：《工业工程与管理》2013年第2期92-99,共8页Industrial Engineering and Management

基　　金：国家自然科学基金资助项目(70871086)

摘　　要：面向诊断相关组(DRGs)的病种成本预测是医院成本管理的重要环节。针对病种成本预测的多因素、非线性特点以及病种系列中医疗服务配置的相似性,提出了一种基于GBOM和改进SVM的DRGs病种系列成本预测方法。首先,提出了基于类物料清单(GBOM)的病种系列成本影响因素表达模型;然后,针对SVM区分样本属性重要性差以及参数选择对预测结果影响较大的问题,提出了一种粗糙集属性(成本影响因素)约简与粒子群算法(PSO)参数寻优相结合的改进SVM病种系列成本预测模型。最后,以阑尾炎病种系列进行了实证研究,预测精度和速度优于BP神经网络、标准SVM和PSO-SVM,进而证明了该方法的有效性和优越性,为病种成本提供了有效的预测方法并显著提高了医院成本控制的准确性。Disease cost estimation for DRGs is an important aspect in hospital cost management. Considering the characteristics of disease cost estimation such as multifactor, nonlinearity and the similarity among various case medical service configurations in the same disease series, this paper proposes the method of disease series cost estimation of DRGs mode based on GBOM and improved SVM. First, this paper proposes the model of disease series cost factors influenced expression based on GBOM（Generie Bill of Material）, and then in allusion to SVM^s defects of distinguishing the importance of attributes and large impacts from parameters selection on estimation results, an improved SVM model of disease series cost estimation based on RS（Rough Sets） for attribute （cost influence factor） reduction and PSO （Particle Swarm Optimization） for parameters optimization has been proposed. Finally, an example of cost estimation for appendicitis disease series is given, and the accuracy and speed of improved SVM estimation model is greater than BP-ANN, the normal SVM and PSO-SVM. The result verifies the validity and superiority of the method, which provides an effective estimation method of disease cost and improves the accuracy of hospital cost control.

关 键 词：诊断相关组 病种系列 成本预测 类物料清单 粗糙集 粒子群算法 支持向量机 

分 类 号：F270[经济管理—企业管理] TP18[经济管理—国民经济]