Melt granulation in fluid bed processors is an emerging technique, but literature data regarding the modeling of this granulation method are lacking. In the present study different techniques (response surface analysis, multilayer perceptron neural network, and partial least squares method) were applied for modeling of spray-on fluidized bed melt granulation. Experiments were organized in line with central composite design. The effect of binder content and spray air pressure on granule properties was evaluated. The results obtained indicate that binder content can be identified as a critical factor controlling the granule size and size distribution. It was found that agglomeration mechanism involved, i.e., granule shape, can be greatly influenced by binder properties. The spray air pressure was identified as critical process parameter affecting granule flowability. The results presented indicate that application of in silico tools enables enhanced understanding and better control of novel pharmaceutical processes, such as melt granulation in fluidized bed. The artificial neural networks and partial least squares method were found to be superior to response surface methodology in prediction of granule properties. According to the results obtained, application of more advanced empirical modeling techniques complementary to design of experiments can be a suitable approach in defining the design space and optimization of spray-on fluidized bed melt granulation.