三层BP神经网络对药品的销售进行预测。：输入层有四个结点，隐含层结点数为5，隐含层的激活函数为tansig；输出层结点数为1个，输出层的激活函数为logsig，并利用此网络对药品的销售量进行预测，预测方法采用滚动预测方式，即用前四个月的销售量来预测第四个月的销售量。-Three-layer BP neural network to forecast the sales of drugs. : Input layer has four nodes, 5 hidden layer nodes, hidden layer activation function tansig　output layer nodes is 1, the output layer activation function is logsig, and use the network of drug sales volume to predict the prediction method using rolling forecasts, that is used to predict sales of the first four months the first four months of sales.