版权所有:内蒙古大学图书馆 技术提供:维普资讯• 智图
内蒙古自治区呼和浩特市赛罕区大学西街235号 邮编: 010021
作者机构:Key Laboratory of Cloud Computing and Intelligent Information Processing of Changzhou City Jiangsu University of Technology Zhongwu Road 1801 Changzhou213001 China Department of Pediatrics Changzhou No.2 People's Hospital Xinglong Road 29 Changzhou213003 China Guangxi Higher Education Key Laboratory of Science Computing and Intelligent Information Processing Guangxi Teachers Education University Yanziling Road 4 Nanning530023 China
出 版 物:《International Journal Bioautomation》 (Int. J. Bioautomotion)
年 卷 期:2015年第19卷第4期
页 面:543-554页
核心收录:
主 题:Momentum
摘 要:Childhood nephrotic syndrome is a chronic disease harmful to growth of children. Scientific and accurate prediction of negative conversion days for children with nephrotic syndrome offers potential benefits for treatment of patients and helps achieve better cure effect. In this study, the improved backpropagation neural network with momentum is used for prediction. Momentum speeds up convergence and maintains the generalization performance of the neural network, and therefore overcomes weaknesses of the standard backpropagation algorithm. The three-tier network structure is constructed. Eight indicators including age, lgG, lgA and lgM, etc. are selected for network inputs. The scientific computing software of MATLAB and its neural network tools are used to create model and predict. The training sample of twenty-eight cases is used to train the neural network. The test sample of six typical cases belonging to six different age groups respectively is used to test the predictive model. The low mean absolute error of predictive results is achieved at 0.83. The experimental results of the small-size sample show that the proposed approach is to some degree applicable for the prediction of negative conversion days of childhood nephrotic syndrome.