The management of spare part has been a crucial thing regarding the availability and reliability of a machine. Managing of spare part inventory is different with managing of material at production process. The forecast of the spare part need with a special demand pattern needs a suitable method. Exponential Smoothing method that was often used giving forecast result that is not suitable with what the spare part really needs. In this present research, the Croston method is used for forecasting spare part demand. This method considers the mean of non-zero demand as well as the interval of it. Moreover, this method is going to be compared to Syntetos-Boylan Approximation (SBA) which is the advancement of the Croston method. Before classifying the spare part based on the analysis of FSN which is going to get the forecast spare part. From the error comparison, SBA method has less error score. It has least score on 14 kind of spare parts from the 16 spare parts, which already predicted. Meanwhile the 2 kind of spare parts has least score on Croston method.