A New Approach for Demand Forecasting and Inventory Optimization of the Rarely Used Spare Parts
-
摘要: 不常用备件具有需求量变化剧烈、需求发生间隔期长且不确定的特点,从而引发了备件需求预测失准而无法做出合理库存决策的问题。为此,提出了一种新的需求预测和库存优化相结合的方法以提高决策的精准度。利用高斯过程回归法预测需求发生间隔,结合Bootstrap增广样本统计方法来预测备件的需求概率分布;基于该概率统计结果,建立库存总成本最小的随机库存优化模型,采用粒子群算法求解得到最优库存决策变量。两个实际工业备件的实验结果表明,本文方法具有更高的预测精度,同时,在满足服务水平约束下库存决策实现了更低的库存总费用,具有实用性。Abstract: The main characteristic of rarely used spare parts include the sharp change of demand, and long and uncertain demand interval, which will result in the inaccurate prediction on the spare part demand such that it is difficult to make a reasonable inventory decision. Aiming at the above issues, this work proposes a novel demand forecasting and inventory optimization method to improve the accuracy of decision-making. In the proposed method, the Gaussian process regression is used to forecast the demand interval, and then, the Bootstrap augmented sample statistical method is combined to predict the probability distribution of spare parts demand. Based on the obtained demand probability statistics results, the stochastic inventory model on the total inventory cost is established and the particle swarm algorithm is further utilized to search the optimal inventory decision variable. Finally, the experimental results from two sets of practical industrial spare parts show that the proposed method has the higher prediction accuracy. Meanwhile, the obtained inventory decision can achieve lower total inventory cost on the premise of satisfying service level, which illustrates the practicality of the proposed prediction and optimization method of infrequent spare partsmethod.
-
表 1 本文方法与其他算法的比较
Table 1. Comparison of the proposed method and other algorithms
Method Data 1 Data 2 MAE RMSE MAE RMSE This paper 0.398 0.208 0.4980 0.4480 ES 0.757 0.787 0.7980 0.8204 SVM 0.900 1.157 0.6630 0.5890 表 2 两种方法的评估结果
Table 2. Evaluation results of two methods
Lead period Actual demamd This paper Literature[10] Confidence interval Safety stock Confidence interval Safety stock 5 2 [0,4] 3 [0,6] 4 8 4 [1,6] 4 [0,8] 6 11 5 [1,7] 6 [1,10] 8 表 3 历史需求量统计信息
Table 3. Historical demand statistics
Demand quantity Occurrence number Frequency/% 0 25 19.53 1 21 16.41 2 33 25.78 3 25 19.53 4 13 10.16 5 3 2.34 6 4 3.13 7 2 1.56 8 1 0.78 9 1 0.78 -
[1] AKPINAR M, YUMUSAK N. Day-ahead natural gas forecasting using nonseasonal exponential smoothing methods[C]//2017 IEEE International Conference on Environment and Electrical Engineering and 2017 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe). Milan: IEEE, 2017: 1-4. [2] KALAVA P, TERMSUKSAWAD P, WASUSRI T. Forecasting lumpy demand for planning inventory: the case of community hospitals in Thailand[C]//2019 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM). Macao: IEEE, 2019: 686-690. [3] LEENATHAM A, KHEMAVUK P. Demand forecasting using artificial neural network based on quantitative and qualitative data[C]//2020 1st International Conference on Big Data Analytics and Practices (IBDAP). Bangkok, Thailand: [s.n.], 2020: 1-6. [4] JIANG A P, KWOK L T, GUO X Y, et al. A new approach to forecasting intermittent based on the mixed zero-truncated poisson model[J]. Journal of Forecasting, 2020, 39(1): 69-83. doi: 10.1002/for.2614 [5] WILLEMAIN T R, SMART C N, SCHWARZ H F. A new approach to forecasting intermittent demand for service parts inventories[J]. International Journal of Forecasting, 2004, 20: 373-387. doi: 10.1016/S0169-2070(04)00020-2 [6] ARIS A, SYNTETOS M, ZIED B, et al. Forecasting intermittent inventory demands: Simple parametric methods vs. bootstrapping[J]. Journal of Business Research, 2015, 68(8): 1746-1752. doi: 10.1016/j.jbusres.2015.03.034 [7] YONIT B, OPHER B. The residual time approach for (Q, r) model under perishability, general lead times, and lost sales[J]. Mathematical Methods of Operations Research, 2020, 92: 601-648. doi: 10.1007/s00186-020-00717-7 [8] WANG Y, SHI Q. Improved dynamic pso-based algorithm for critical spare parts supply optimization under (T, S) inventory policy[J]. IEEE Access, 2019, 7: 153694-153709. doi: 10.1109/ACCESS.2019.2948859 [9] EHRENTHAL J C F, HONHON D, VAN WOENSEL T. Demand seasonality in retail inventory management[J]. European Journal of Operational Research, 2014, 238(2): 527-539. doi: 10.1016/j.ejor.2014.03.030 [10] 刘琴, 申海, 朱美琳. 基于需求分布规律的石化企业备件库存优化研究[J]. 数学的实践与认识, 2020, 50(9): 116-121. [11] LAIB O, KHADIR M T, MIHAVLOVA L. A Gaussian process regression for natural gas consumption prediction based on time series data[C]//2018 21st International Conference on Information Fusion (FUSION). Cambridge UK: IEEE, 2018: 55-61. [12] GU JY, ZHANG G Q, LI K W. Efficient aircraft spare parts inventory management under demand uncertainty[J]. Journal of Air Transport Management, 2015, 42: 101-109. doi: 10.1016/j.jairtraman.2014.09.006 [13] TALEIZADEH ATA ALLAH, TAFAKKORI KEIVAN, THAICHON PARK. Resilience toward supply disruptions: A stochastic inventory control model with partial backordering under the base stock policy[J]. Journal of Retailing and Consumer Services, 2021, 58: 102291. doi: 10.1016/j.jretconser.2020.102291 [14] WANXIA M, SHIHUA M. Optimal control of (q, r) Inventory system with service level constraint[C]//2010 International Conference on Computing, Control and Industrial Engineering, Wuhan: IEEE, 2010: 401-404. [15] 王玮. 集成多SVM的不常用备件需求预测支持系统研究[D]. 武汉: 华中科技大学, 2007: 44-45. [16] 宋铭星. 动车组不常用备件的判定以及需求预测研究[D]. 北京: 北京交通大学, 2017: 36-37. [17] BOUKHTOUTA A, JENTSCH P. Support vector machine for demand forecasting of canadian armed forces spare parts[C]// 2018 6th International Symposium on Computational and Business Intelligence (ISCBI), Basel, Switzerland: IEEE, 2018: 59-64. [18] 王学武, 闵永, 顾幸生. 基于密度聚类的多目标粒子群优化算法[J]. 华东理工大学学报(自然科学版), 2019, 45(3): 449-457. -