Application of Wavelet Neural Network to the Prediction of Gas Chromatographic Retention Indices of Alkanes
Chunsheng Yin, Weimin Guo, Teng Lin, Shu‐Shen Liu, Rongqiang Fu, Zhongxiao Pan, Liansheng Wang · Journal of the Chinese Chemical Society · 2001
Abstract A wavelet neural network (WNN) is employed to create a quantitative structure‐retention index relationship, which correlates the novel molecular distance edge vector (MDEV)‐consisting of ten elements to Gas Chromatographic retention indexes (RIGC) of Alkanes. The RIGC has been calculated by the WNN from the molecular topological descriptors of examined alkanes. In this work, the RIGC estimated and predicted by conventional neural networks (say back propagation neural networks, BP) has also been provided. The excellent predicted results with a correlation of 0.9996 and standard deviation of 5.0598 suggest that the WNN technique is a powerful tool in QSAR/QSPR modeling and superior to the BP neural networks.