摘要
使用KPCA(核主成分分析)对含能化合物的结构参数进行参数选择,在保持原有数据主要信息的情形下,得到数据的主成分。将降维后的特征信息作为GRNN(广义回归神经网络)的输入,含能化合物的性能数据作为输出,建立非线性的定量含能化合物结构性能关系预测模型。与PCA_GRNN模型的比较表明,该模型能很好地反映含能化合物结构和性能之间的关系,具有较高的预测正确率。
In this article it used. KPCA （ kernel principal components analysis） to select parameter for the structure parameters of energetic compounds, and the principal components of the data is obtained in the case of maintaining main information of original data. Then these principal components were used as the inputs of the GRNN （ Generalized Regression Neural Networks） after dimensions reduction and the performance data of energetic compounds were used as outputs, all for constructing a nonlinear prediction model of structural performance relationship of quantitative energetic compounds. The comparison result with PCA_GRNN model indicated that this model can well reflect the relation between structures and performances of energetic compounds and has the high prediction precision.
出处
《计算机应用与软件》
CSCD
2009年第7期112-114,共3页
Computer Applications and Software
基金
陕西省自然科学基金(98X11)
陕西省教育厅重点科研计划项目(00JK015)
