基于聚类和支持向量机的胃癌患者住院费用建模（英文）

发布时间：2018-10-08 18:54

【摘要】：针对胃癌患者住院费用分类标签设定的复杂性以及传统费用建模算法的局限性,本文提出了一种基于聚类和支持向量机的住院费用建模算法,为胃癌患者住院费用的控制和预测提供方法基础.搜集整理宁夏某三甲医院2009 2011年间1583例胃癌患者为样本,采用K-means对总住院费用逐年聚类得到分类标签,最后通过支持向量机对住院费用进行建模预测以及影响因素分析,用分类准确率作为预测效果的评价指标.实验结果表明胃癌患者住院费用呈逐年增加趋势,其中以西药费为主,占总费用的53.74%.通过K-Means以年份对费用聚类比单纯以费用分布特征聚类的分类准确率提高了13.13%,当核函数选用高斯核函数,且惩罚因子C=10和核参数γ=1时建立的支持向量机模型最稳定,分类准确率为92.11%.实验结果表明根据年份聚类得到类别标签更合理,结合聚类的SVM来预测住院费用更有效.
[Abstract]:In view of the complexity of the classification label setting of hospitalization expenses of gastric cancer patients and the limitation of the traditional cost modeling algorithm, this paper proposes a hospital cost modeling algorithm based on clustering and support vector machine. To provide a method basis for the control and prediction of hospitalization cost of gastric cancer patients. This paper collected and sorted out 1583 gastric cancer patients from 2009 to 2011 in a third Class A Hospital in Ningxia, and used K-means to cluster the total hospitalization cost year by year to get the classification label. Finally, the cost of hospitalization was modeled and predicted by support vector machine and the influencing factors were analyzed. The accuracy of classification is used as the evaluation index of prediction effect. The results showed that the hospitalization cost of gastric cancer patients was increasing year by year, in which western medicine was the main expense, accounting for 53.74% of the total cost. The classification accuracy of cost clustering based on K-Means is improved by 13.13%. When the kernel function is Gao Si kernel function, the support vector machine model established by penalty factor Cn10 and kernel parameter 纬 = 1 is the most stable. The classification accuracy is 92.11%. The experimental results show that it is more reasonable to get the category label according to the year clustering, and it is more effective to predict the hospitalization cost by combining the clustering SVM.
【作者单位】：宁夏医科大学理学院;
【基金】：Supported by National Natural Science Foundation of China(61561040) Natural Science Foundation of Ningxia(NZ16067) Natural Science Foundation of Ningxia Education Department(NGY2016084)
【分类号】：R197.323;TP18

【相似文献】