2型糖尿病患病因素对血糖影响的定量分析

为了解血糖变化特点和确定2型糖尿病干预靶点，提出了一种基于反向传播（BP）神经网络的患病因素对血糖影响的量化分析方法. 筛选出性别、体重等10维因素，再根据性别和年龄将人群分成9组，分别训练BP模型并提取敏感度，得出因素对血糖影响的量化值. 实验结果表明，体重对血糖变化的敏感度最大（0.244 9），体重、血脂水平和年龄敏感度各占总量的25%、35%和17%，且不同性别、年龄敏感度排序不同，为个体细化干预提供了判定方法；该方法可推广到其它疾病危险因素的量化分析，还可用于因素干预-判定-因素干预的良性循环中，有效提升个体健康水平. 

Quantitative Analysis of Type 2 Diabetes Risk Factors Effect on Blood Glucose Change

To understand the characteristic of blood glucose change and confirm intervention points of type 2 diabetes, a quantitative method was proposed to analyze the effect of risk factors on blood glucose change based on BP neural network. Firsly ten risk factors were screened. Then the cohort was divided into nine groups by sex and age. Nine BP neural network models were trained respectively and quantitative values of the effect of different risk factors on blood glucose change could be obtained by sensitivity calculation. The experiment results indicate that the sensitivity of weight related to blood glucose change is 0.244 9 which is the largest value. The sensitivity ratios of weight, blood lipid level and age are 25%, 35% and 17% in the total respectively. Moreover, the sensitivity sequences are different in different sex and age groups, which can supply judgment method for individual intervention. This method can promote to risk factors quantitative analysis of other diseases and use in the virtuous circle of factors intervention-judgment-factors intervention to improve the individual health.