摘要
结合案例分析,将神经网络方法应用于医学统计预测,并与传统的回归分析进行了预测效能的比较。结果表明,神经网络方法与回归分析在分类效能上比较相近,但神经网络方法的容错性更好,能通过训练和学习获得完整的预测规则。在本研究中,对于连续变量的处理,多层感知器表现出比径向基神经网络更占优势,而在对离散变量进行分类时,径向基神经网络的结果则更为优良。总之,神经网络方法在医学统计中可以获得有效应用,并能预测比回归分析更多的结果,因而在大数据时代具有更广阔的应用前景。
Neural network method was applied to medical statistics forecast with the combination of case analysis, with its forecast efficiency being compared to that of traditional regression analysis. It was testified that neural network method has similar classification efficiency with regression analysis, while it had better fault-tolerance property. Hence, the integral predict rules can be obtained through training and learning. In this study, multilayer perceptron showed advantages over radial basis function neural network in dealing with continuous variables. While the radial basis function neural network was better at dealing with the discrete variables. In a word, neural network method has been efficiently used in medical statistics, with more results being detected than regression analysis, which therefore enjoys a broader application future in the big date era.
出处
《南京中医药大学学报（社会科学版）》
2017年第1期47-52,共6页
Journal of Nanjing University of Traditional Chinese Medicine(Social Science Edition)
关键词
神经网络
回归分析
医学统计
neural network
regression analysis
medical statistics
