基于人工神经网络的碱性蛋白酶水解乳清蛋白预测模型的建立

The model building of whey protein hydrolyzate by alcalase based on artificial neural network

摘要: 乳清蛋白水解物的生物利用价值高,水解度是衡量水解物的重要指标之一。基于人工神经网络方法建立了碱性蛋白酶水解乳清蛋白的水解模型。利用Matlab的nnet工具包中的nftool建立神经网络模型。实验中设置了15个隐藏神经元,选用Levenberg-Marquardt训练算法迭代计算。实验结果表明:测试值与输出值的误差平均较小,拟合回归得到的整体R2=0.97995,测试集合的R2=0.96239,表明神经网络拟合效果和预测能力良好,能够较好地完成水解度预测的目标,选用的测试组的实验结果与神经网络输出值之间的误差均在±3%以内,说明此神经网络预测模型具有较高的预测能力和精度。

Abstract: The whey protein hydrolyzate with high value bioacailability, the degree of hydrolysis ( DH%) is an important indicator to evaluate the hydrolysate.An artificial neural network ( ANN) model was used to predict whey protein hydrolysis process.Analyses performed on a nftool in the nnet kit of Matlab.15 hidden neurons were set up, and the Levenberg-Marquardt was chosen for training.The results showed the error between test value and output value was slight, fitting regression were overall R2= 0.97995, the test set R2= 0.96239, this indicated that the effect and predictive ability of the model was good, error between experimental and calculated was less 3%, this model had a high predictive ability and precision.