人工神经网络预测别嘌醇HP MC缓释片药物释放

目的用人工神经网络模型定量的预测HPMC的量和其固有黏度对药物释放的影响。方法以难溶性药物别嘌醇为模型药物,固定其他因素,HPMC的量和HPMC的固有黏度作为自变量,设计了18个处方并进行释放度检查;其中的13个处方作为训练处方,其他5个处方为验证处方,将上述的变量作为人工神经的输入,以药物在各个取样时间点的释放为输出,采用剔除一点交叉验证法建立人工神经网络模型。通过线性回归和相似因子说明人工神经网络的预测能力。结果训练和验证处方人工神经网络预测值与实际测定相符。结论建立BP人工神经网络,根据HPMC的量和其固有黏度可以定量的预测药物在各个时间点的药物释放。

Using Artificial Neural Network (ANN) to quantitatively predict allopurinol release from HPMC sustained release tablets

Objective To use ANN to quantitatively predict the influence of the amount of HPMC and its internal viscosity on allopurinol release from HPMC sustained release tablets.Methods The insoluble drug allopurinol was used as model drug,and the amount of HPMC and internal viscosity of HPMC were used as casual factors(inputs),while drug accumulating drug rel ease in each sampling time was used as response variables(outputs).18 formulations were prepared,13 of the 18 formulations were used as training data sets,and the other 5 formulations were used as validating formulations."leave one out" cross validation was used as training approach for the ANN.The regression equations and seminary factors were used to judge the ability of the ANN.Results The linary regression equation parameter and the seminary factors show that,for both