Refrigeration processes are present in many areas of everyday life, from air conditioning and residential refrigerators to applications involving the preservation of products with high added value. The need to develop other control strategies may arise and having a model representing the dynamics characteristics of the system can facilitate this task by reducing costs and time of simulation. In this work is sparse models ARX and ARMAX for prediction evaporating temperature of a refrigeration system using the FROLS algorithm for determining the ERR and AOE performance criteria. The developed models have good performance with AOE 0.0419 °C for model ARX model and 0.0250 °C for the model ARMAX.