论文部分内容阅读
为研究野韭菜挥发性成分的性质,预测其色谱保留指数,运用MATLAB相关自编程序计算得到了野韭菜挥发性成分的分子形状指数和电性拓扑态指数,将这两类参数作为分子结构描述参数,借助多元逐步回归法优化筛选了其中结构参数2K、3K、4K、I2和I6,建立了野韭菜挥发性成分色谱保留指数的QSRR模型,相关系数为0.963,通过对模型的稳定性和预测能力进行检验,检验的相关系数r也稳定在0.963左右。用这5个筛选出的结构参数作为人工神经网络的输入层参数,采用5-2-1的网络神经结构,利用BP算法建构神经网络模型,总相关系数达到0.996的优级相关,利用此模型计算得到的预测值与实验值吻合度较为理想,相对平均误差仅为1.67%,结果显示BP神经网络所得结果优于多元线性回归方法。
In order to study the properties of wild chives volatile components and predict the chromatographic retention index, the molecular shape index and electrical topological index of volatile components in wild chives were calculated by MATLAB self-programming program. The two types of parameters were described as molecular structure QSRR model was established for the determination of the chromatographic retention index of the volatile components in wild Chinese chives by using multiple stepwise regression to optimize the model parameters 2K, 3K, 4K, I2 and I6. The correlation coefficient was 0.963. Through the stability and prediction Ability to test, test the correlation coefficient r also stabilized at about 0.963. Using the five selected structural parameters as the input layer parameters of artificial neural network, the neural network model of 5-2-1 was adopted and the neural network model was constructed by BP algorithm. The total correlation coefficient reached 0.996, The calculated predictive value is in good agreement with the experimental value, and the relative average error is only 1.67%. The result shows that BP neural network is better than multivariate linear regression.