Abstract: Accurate demand forecasting of the high-priced and repairable line replaceable unit (LRU) parts is an important basis for airline procurement and an important way to reduce costs. Taking the Engine Driven Pump as an example, traditional forecasting models and the common measurement forecasting models for aviation spare parts were selected to forecast demand; then combined with the comparison of evaluating indexes, the best model was obtained; finally, the forecast demand of the models were discussed and compared with the practices to verify the conclusion. The results show that the negative binomial regression model has obvious advantages among the six models for the demand forecast of LRU parts. It has the lowest AIC, 217.0601. The prediction errors in 2018 and 2019 are only 0.1693 pieces and 7.3850 pieces, and it can meet the realistic requirements of airline flight guarantee rates exceeding 95%. In the case of only the number of failures is available, the measurement forecasting models are more advantageous. The comparison results of the forecasting models can provide reference for the airline's decision to purchase LRU parts.

Key words: demand forecasting, comparative analysis, regression analysis, civil aviation, line replaceable unit (LRU)