Short period undulators, producing high magnetic fields, are an essential component in the production of x-rays in compact free electron lasers (FELs) and medium energy storage facilities. A short-period undulator can be built in different ways, however, the staggered array undulator geometry stands out for its simplicity in design and manufacturing efforts. For the operation of a staggered array type undulator, the superconducting material (either ReBCO bulks or stacks of 2GReBCO tapes) has to be magnetized in an external magnetic field in order to produce the desired undulator field. Hereby it can be difficult to predict the correlation between the applied external field and the resulting undulator field. In this publication we a present a possible approach for predicting this complex magnetization process and the consequential undulator field with the help of an artificial neural network. This prediction method can yield results in real time and would therefore be ideal to be used in the control loop of a FEL or a storage ring.