基于EFDT-HS随机森林算法的货物装载研究  

Multi-specification cargo loading based on EFDT-HS random forest algorithm

作　　者：李伟 杨超宇[1] 孟祥瑞[1] LI Wei;YANG Chao-yu;MENG Xiang-rui(School of Economics and Management,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学经济与管理学院,安徽淮南232001

出　　处：《哈尔滨商业大学学报（自然科学版）》2020年第4期500-505,512,共7页Journal of Harbin University of Commerce：Natural Sciences Edition

基　　金：国家自然科学基金(61873004,51874003);安徽教育厅人文社会科学研究项目(SK2017A0098);安徽理工大学博士基金(11892).

摘　　要：为保证多规格货物在集装箱中的高效装载,提出了一种启发式随机森林算法,该算法首先利用Bagging方法生成T个训练集,对每个训练集,要求在特征集中随机选取K个特征组成新的特征集,将新特征集中的最优特征作为分割特征,利用分割特征计算并择优选取样本信息熵,然后构建生成货物装箱决策树模型,最后基于启发式搜索方法对货物装载后的剩余空间进行合并再利用.通过BR1~BR10共十组异构性逐渐增强的货物数据对算法进行仿真实验,将实验结果与其他研究算法进行比较,该算法在BR10算例中利用率达到90%,仅比其他算法低1%,但计算时间却由138062 s降低到76 s.由此可见,该算法对于多规格强异构货物的求解具有一定的可行性和有效性.In order to ensure the efficient loading of the goods in the container of many specifications,this paper proposed a heuristic random forest algorithm,this algorithm firstT generated by Bagging method a training set,for each training set,the requirements on the characteristics of concentrated randomKfeatures of new set,the optimal features as separate from the new collection,using information entropy segmentation feature for calculation and optimal selection samples,and build produces goods the decision tree model,using the last method based on heuristic search after loading the goods of the remaining space to merge.Finally,ten groups of goods data with gradually increasing heterogeneity of BR1~BR10 were used to carry out simulation experiments on the algorithm,and the experimental results were compared with other research algorithms.The utilization rate of the algorithm reached 90%in the BR10 example,only 1%lower than other algorithms,but the calculation time was reduced from 138062 s to 76 s.It can be seen that this algorithm has certain feasibility and effectiveness for the solution of multi-specification strongly heterogeneous goods.

关 键 词：集装箱装载 随机森林 启发式搜索 极快决策树 空间优化 多规格货物 

分 类 号：U169[交通运输工程]