基于神经网络的高血压在线风险评估系统

由于先前的研究工作不够综合和精确,不足于建立准确的高血压风险评估系统,根据2231个正常样本及823个高血压样本计算的信息增益,对高血压致病因素的重要程度进行了排序,总共建立和测试了42个不同的神经网络模型,发现了一个输入为26个致病因素的神经网络模型,其预测精度远高于先前研究取得的81.61%,该模型关于“是否高血压”、“收缩压”、“舒张压”的预测符合率分别为95.79%,98.22%和98.41%.基于发现的神经网络模型及面向对象的技术,开发了一个能自动收集新样本、学习新样本并能改进预测精度的高血压风险在线评估系统.

Neural network based online hypertension risk evaluation system

Since the previous research works are not synthetic and accurate enough for building a precise hypertension risk evaluation system,by ranking the significances for hypertension factors according to the information gains on 2 231 normotensive and 823 hypertensive samples,totally 42 different neural network models are built and tested.The prediction accuracy of a model whose inputs are 26 factors is found to be much higher than the 81.61% obtained by previous research work. The prediction matching rates of the model for "hypertension or not","systolic blood pressure",and "diastolic blood pressure" are 95.79%,98.22% and 98.41%,respectively.Based on the found model and the object oriented techniques,an online hypertension risk evaluation system is developed,being able to gather new samples,learn the new samples,and improve its prediction accuracy automatically.