基于梯度下降灰色Markov模型的装备器材需求预测  

Material Demand Forecasting Based on Gradient Descent Grey Markov Model

作　　者：张磊[1] 李世民 康淑瑰 王铁宁 ZHANG Lei;LI Shimin;KANG Shugui;WANG Tiening(Shanxi Datong University,Datong 037009,China;Unit 63963 of PLA,Beijing 100072,China;Army Academy of Armored Force,Beijing 100072,China)

机构地区：[1]山西大同大学,山西大同037009 [2]解放军63963部队,北京100072 [3]陆军装甲兵学院,北京100072

出　　处：《火力与指挥控制》2022年第5期53-59,共7页Fire Control & Command Control

基　　金：国家自然科学基金(11871314);山西省哲学社会科学规划基金资助项目(2021YY203)。

摘　　要：为改进灰色Markov装备维修器材需求预测模型在小样本条件下的预测效果与合理性,提出了基于梯度下降算法的灰色Markov模型。通过建立关于各时间节点拟合残差的二乘损失函数,运用梯度下降算法迭代得到灰色Markov模型最优转移概率强度和Kolmogorov方程解的最优待定系数并实施预测。数值模拟与实例分析的结果表明,梯度下降灰色Markov模型可快速有效地搜索转移概率强度以及待定系数的相对优化值,进而改善模型的预测效果和稳定性。In order to improve the prediction effect and rationality of grey Markov model which is used to predict equipment maintenance material with small sample,Gradient Descent Grey Markov model is proposed in this paper. The optimal transfer probability intensity and undetermined coefficient of Kolmogorov equation are obtained by using gradient descent method,then predictive result is obtained. The results of numerical simulation and case analysis show that the gradient descent Grey Markov model can quickly and effectively search the relative optimal value of the transition probability intensity and undetermined coefficient,and then the prediction accuracy and stability of the model are improved.

关 键 词：器材需求预测 灰色Markov GM(1 1) 梯度下降算法 

分 类 号：E92[军事—军事装备学] TJ01[兵器科学与技术—武器系统与运用工程]