摘要
目的:考察数据挖掘技术对肝癌患者术后预测的效果。方法:采用数据挖掘技术中的神经网络方法在SAS软件上构建神经网络模型。利用该模型分析我校东方肝胆外科医院1990年1月1日入院至1995年12月31日出院的1457例肝癌患者的临床资料,检测其对肝癌患者术后复发的预测准确率。结果:从56个指标中筛选出有统计学意义的11个指标进入模型,经适当训练后在验证集和测试集上预测准确率均在80％以上。结论:数据挖掘技术可应用于肝癌术后预测分析,效果较为理想。在应用数据挖掘技术时应采用较大样本,同时还应根据资料类型选用适当的方法,反复尝试。
Objective:To examine the effectiveness of data mining in postoperative prediction of liver cancer. Methods: The neural network model was constructed with SAS software by applying the method of neural network. With this model, clinical data of 1 457 patients in Eastern Hepatobiliary Surgery Hospital from Jan. 1,1990 to Dec. 31, 1995 were analyzed to examine the prediction accuracy of data mining neural network for postoperative recurrence of liver cancer. Results: Eleven statistically significant indices were selected from 56 for the model. After appropriate training, the prediction accuracy was above 80% both on validation set and test set. Conclusion:Data mining can be applied in postoperative prediction analysis of liver cancer and the effect is satisfactory. Big samples should be adopted in data mining, and proper method should be chosen according to data type and should be tried repeatedly.
出处
《第二军医大学学报》
CAS
CSCD
北大核心
2003年第11期1241-1243,共3页
Academic Journal of Second Military Medical University
基金
国家自然科学基金(39770835)
关键词
数据挖掘技术
肝癌
术后预测
术后复发
预后
data mining
neural network
prognosis
prediction analysis
hepatoma
postoperative recurrence
