基于BP神经网络的我国石油需求预测

Forecast of oil consumption based on BP neural network

具有Sigmoid型隐层函数的3层BP神经网络能够以任意精度逼近任何一个具有有限间断点的非线性函数．所以，选择Sigmoid函数的3层Bp神经网络作为石油消费预测模型．根据我国1989-2002年国民生产、石油消费数据进行数据训练学习，以2003年国民生产数据作为输入，预测2004年石油消费．结果表明，该模型预测精度较高，误差仅为1．87％，具有较好的应用价值．

A three-ply BP neural network with sigmoid function of layers can approximate any nonlinear functions with finite point of discontinuity in arbitrary precision. So we use the three-ply neural network with sigmoid function of layers as the model to forecast the oil consumption. This paper makes the forecast of the oil consumption in 2004 by training the neural network through the economic data and oil consumption from 1989 to 2002 and inputting the data of 2003 to forecast the oil consumption of 2004. Compared with the real data of 2004, the error of the forecast is 1.87%. The precision of the forecast proves that the method is highly applicable.