The Production Quality. Medicine Based On Bp Neural Network Prediction And Applications

Now Medicine Production company's main method is to produce medicine first, then to test the medicine's quality。 To forecast the medicine' s quality beforehand use manual work mode, this method has some difficulty in dealing with multi-factor quality general forecast, especially in quantitative analysis。 BP Neural Network is a good machine learning arithmetic, it has some advantage in classification,function proximity, forecast etc, it is apt to solve nonlinear difficulty problems , so this article use BP Neural Network to solve Medicine Production quality forecast problems。This article first analyze the theory of BP Neural Network, then put forward a production quality forecast model based on BP Neural Network, and introduce the process of model founding in detail , including forecast model character's choice , data preparation for training and test, realization of BP Neural Network in MATLAB software 。 This article take eliminating poison benefit for the lung fluid medicine in eating as an example, after the analysis of the primary factors of influences on quality, the input character of the quality forecast model and the output factors that control the medicine quality has been chosen, and use MATLAB software to establish the BP Neural Network model in quality forecast, and tested the model with simulated data。In the BP Neural Network founding method, two ways have been used in experiments: one is to use the character vector data directly in establish the Neural Network, the other is to switch the Character Vector data into [-1, 1] first, then establish the model。 To put up founding and test experiments in these two ways separately, the first model use fixed training data to train the network and it constringed at the ninety circles, it's test identifying rate is eighty eight percent;the second model use thesame fixed training data to train the network and it constringed at the sixty circles, it's test identifying rate is ninety nine percento Thus shows that the model with the switched character data is more better in identifying rate and it' s training time is shorter than the model used the character data directly,, Use BP Neural Network to forecast the Medicine Production quality is a better method..