基于Clementine的毛细管电泳数据挖掘方法研究

摘要

To predict capillary eleetrophoresis mobility time, Clementine' s neural network model is used to predict the eleetrophoresis mobility of 7 drugs of vitamin B1, etc. 350, as the 50 sets of data for the drug, taken as learning set (or training set) of 49 groups for training, and then takes the 1 set of data as prediction set. The most of the prediction results of relative error of Clementine is less than 10％, the prediction result is more accurate, for large sample data using neural network model of data mining to predict will get better results.%为预测毛细管电泳迁移时间,利用数据挖掘软件Clementine的神经网络模型对维生素B1等7种药物的350个迁移时间数据进行预测,对于药物的50组数据,取49组作为学习集(或称训练集)进行训练,然后用训练好的网络对剩下的1组进行预测.Clementine的预测结果相对误差大部分小于10%,预测结果比较准确,对于大样本数据利用数据挖掘中的神经网络模型进行预测会得到比较好的结果.