The critical spare parts (CSP) is essential to machine operation, which is also more expensive, have longer purchasing lead time and larger demand variation than non-critical spare parts. When the equipment is operating, critical spare parts required to be changed due to wear and tear. Excessive critical spare parts will cause accumulation of the inventory and insufficiency will cause termination of machine operation, thereby leading to loss. Therefore, it is an important issue to devise a way to forecast the future required amount of CSP accurately.    This investigation applied grey prediction model, back-propagation network and moving average method to forecast the CSP requirement in a semiconductor factory, so as to effectively predict the required number of CSP, which can be provide as a reference of critical spare parts control.