Single-input-single-output (SISO) and multi-input-single-output (MISO) models of a cooling crystallization process are developed using time series and neural network approaches. The process description is simulated using first principles including the population balance to describe the crystal size distribution. The process model is identified from the input-output data of the process simulation using either a linear ARX model or a non-linear model constructed using a recurrent or a feedforward neural network.