基于PSO与ELM组合算法的装备器材消耗预测模型  

Equipment Material Consumption Prediction Model Based on the Combined Algorithm of PSO and ELM

作　　者：刘畅[1,2] 伍洁 肖斌 张东东[1] LIU Chang;WU Jie;XIAO Bin;ZHANG Dongdong(Department of Management Engineering and Equipment Economics,Naval University of Engineering,Wuhan 430033,China;不详)

机构地区：[1]海军工程大学管理工程与装备经济系,湖北武汉430033 [2]武警第二机动总队,福建福州350200

出　　处：《武汉理工大学学报（信息与管理工程版）》2022年第1期112-116,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家自然科学基金项目(61802425).

摘　　要：针对装备器材消耗的随机性与波动性而难以准确预测消耗变化的情况,构建了粒子群算法(particle swarm optimization,PSO)和极限学习机(extreme learning machine,ELM)的预测模型。首先,构建训练集和ELM网络;其次,采取PSO优化ELM的输入层权值和隐藏层阈值,解决因权值和阈值随机出现而造成的网络不稳定现象;最后,通过使用3种非线性、非平稳数据集测试及装备器材消耗数据的预测分析,验证了PSO-ELM的预测精度优于EMD-GA-BP、LSSVM、PSO-LSSVM、ELM模型。In view of the randomness and volatility of equipment material consumption,it is difficult to accurately predict future consumption changes.To this end,this paper constructs particle swarm optimization(PSO)and extreme learning machine(ELM)prediction models.First,the training set and ELM network are constructed,and then PSO is used to optimize the input layer weights and hidden layer thresholds of ELM to solve the network instability caused by the random appearance of weights and thresholds.Finally,through the use of three stock index data sets testing and equipment materials consumption data prediction analysis,it is verified that the prediction accuracy of PSO-ELM is better than EMD-GA-BP,LSSVM,PSO-LSSVM,ELM models.

关 键 词：器材消耗 粒子群算法 极限学习机 预测精度 

分 类 号：E917[军事] TJ02[兵器科学与技术—兵器发射理论与技术]