用Boosting算法预测多硝基芳香族化合物的密度

Prediction on Densities of Polynitroaromatic Compounds via Boosting Algorithm

采用Boosting算法对多硝基芳香族化合物(PNACs)的密度进行预估。选用分子结构描述码作为输入特征参数。结果表明,PNACs的密度与其分子结构存在良好的相关性,与人工神经网络相比,Boosting算法对预测的准确性有显著提高,预测结果的相对误差都在8%以内。

The densities of polynitroaromatic compounds （PNACs） are predicted by Boosting algorithm. The molecular structure describers （MSD） are used as input feature parameters. The results show a better correlation between the densities and molecular structures of PNACs. Compared with artificial neural network, the Boosting algorithm greatly improves the prediction accuracy with the relative errors of predicted results within 8%.