Demand forecasting for rush repair spare parts of power equipment using fuzzy c-means clustering and the fuzzy decision tree

Rush repair spare parts of power equipment are necessary for ensuring the normal operation of the critical equipment of power grid. Once the power equipment is damaged due to disasters and the rush repair spare parts are out of stock, the reper-cussions of the accident will be worsened, causing further harm to the power grid and potentially large losses to production and life. On the other hand, the excessive inventory of rush repair spare parts will cause a waste of cost. Therefore, accurate prediction of rush repair spare parts demand is very important. It is particularly difficult to forecast the demand for rush repair spare parts of power equipment due to the nature of their demand, which is usually highly uncertain, random, and with a small amount of historic data. Aiming to improve the scientificity and practicality, this paper proposes a demand forecasting method of rush repair spare parts by using the Fuzzy C-Means (FCM) clus-tering and the Fuzzy Decision Tree (FDT). At first, we analyze the characteristics of emergency events causing spare parts demand and the attributes of spare parts' demand data. Secondly, FCM is applied to dividing the historic demand data into clusters. Then, FDT is used to mine the correlation between the spare parts demand and the emergency events and forecast the demand for rush repair spare parts according to the best cluster of data, and thus realize to predict the demand even with a small data set. Finally, we verify the proposed method by a data set of emergency demand for rush repair spare parts during the last three years from a power company.