Prediction is an important and difficult target during the drilling optimization process. Drawn upon the field data(comprising of a total of 28 types in 4 categories), a ROP prediction model is developed with combination of formation composition and drilling parameters. In order to increase the prediction accuracy and avoid the potential impact caused by its dimension and order of magnitude, the raw data is standardized. In addition, due to the excessive volume of the raw data, it is processed with polymer analysis and factor analysis to pick effective data, and then reduced to a prediction model of 11 input parameters and 1 output parameter. The technology of artificial neural network is used to perform the non-linearly fitting work, and the results indicate that the prediction error is less than 10%, which means the proposed prediction model can somehow provide guidance for field operations.