Coke is an indispensable material in Ironmaking process by blast furnace. To provide good and constant quality coke for stable and efficient blast furance operation is very important. Furthermore, a challenging issue in the cokemaking process is the prediction of coke quality. An accurate prediction can support production planning decision and reduce business operation costs. The objective of this thesis is to apply the backpropagation neural network and the model tree techniques for predicting the strength and meansize of coke. Specifically, we developed the coke- physical&chemical-property model, coal-usage model, coal-group-usage model, and extended model for the target prediction task. Experimentally, we found that the coal-usage model achieves the highest Correlation Coefficient and the lowest Mean Absolute Error. Moreover, the model trees technique reaches higher accuracy and better efficiency than does the backpropagation neural network technique.