MPG Prediction based on BP Neural Network

Abstract

In this article, we use the data mining theory to construct a BP neural network model to predict MPG (mile per gallon). Based on the nonlinear properties of the six variables given, considering the imperfection using two main variables at the same time, we've processed the problem via data preparation, model selection, construction, modification and error comparison as well as model adaption period. At the end of this article, we've discussed the principles of acquiring proper parameters based on the distinctions of the neural network chosen and give some possible improving directions. In this manner, in case the original data is given, the predicted MPG result comes out automatically and satisfactorily