在本論文中使用一組四個參數的描述因子，包括：最低未佔據分子軌道的能量、分子平均極化率、定容比熱和氫原子中最大正電荷來估計高分子的折射率，並且以逆傳遞類神經網路來建造折射率的估計模型。在本文中建立了以4-8-1層架構的最佳類神經網路，經過驗證組的驗證測試，分別得到了均方根誤差為0.0196 (R2 =0.950, Data Ⅰ)與0.0153 (R2 =0.927, Data Ⅱ)的結果。從測試的結果顯示，以類神經網路模式估計高分子的折射率可以得到非常準確的估計值並提供聚合物分子設計的指引。經與其它相關文獻比較，本論文所提的類神經網路模式估計折射率的表現較為優異。
In this article, a set of four-parameter descriptors, including the energy of the lowest unoccupied molecular orbital, the molecular average polarizability, the heat capacity at constant volume, and the most positive net atomic charge on hydrogen atoms in a molecular, were used to correlate with refractive index for polymers. Backpropagation artificial neural network (ANN) were used to generate the models. The final optimum neural network with 4–8–1 structure produced the validation set root mean square errors(RMSEs) of 0.0196 (R2 =0.950, for Data Ⅰ) and 0.0153 (R2 =0.927, for Data Ⅱ), respectively. It suggests that the model obtained here can predict the refractive index values of polymers and provide theoretical guidance for polymeric molecular designs. Compared with the existing ANN models, the proposed model is most accurate in the estimation of refractive index values for polymers.
