基于机器学习的成品卷烟销售订单配送调度 优化算法研究与应用

Research and application of optimization algorithm based on machine learning for scheduling cigarette product delivery

为解决卷烟工业企业成品卷烟销售订单配送调度过程中存在的工作效率低、管理不科学、人工作业量大等问题,基于机器学习设计了一种智能配送调度算法。首先将待调度的成品卷烟销售订单进行凝聚聚类,然后利用专家策略函数和机器策略函数构建搜索树,再通过构造调度优化评价函数进行配送调度决策。以红云红河烟草(集团)有限责任公司2020年4月部分成品卷烟销售订单为例,对精确算法、遗传算法以及机器学习算法进行对比测试。结果表明:与精确算法和遗传算法相比较,机器学习算法具有更好决策性能及全局收敛性,在相同待调度订单数据前提下,使用车辆数分别减少13.28%和10.14%,运输里程减少5%和7.2%,时间成本减少68%和45%,多点组合车次数分别增加11.75和8.07百分点。该方法可为提高成品卷烟销售订单配送调度效率提供支持。

For promoting the efficiency and reducing the labor intensity in scheduling cigarette product delivery,an intelligent scheduling algorithm based on machine learning was designed.First,the sales orders of cigarette products were clustered with the agglomerative cluster.Second,a search tree was constructed by using an expert strategy function and a machine strategy function.Finally,an optimized schedule evaluation function was established and the schedule decision of cigarette product delivery was made.Accurate algorithm,genetic algorithm and the machine learning-based algorithm were comparatively tested by taking part of the sales orders of cigarette products of HongyunHonghe Group in April 2020 as an example.The results showed that the machine learning-based algorithm had better decision making performance and global convergence;for the same sales orders,it used 13.28%and 10.14%less vehicles,reduced transport mileage by 5%and 7.2%,reduced time cost by 68%and 45%,and the number of vehicles serving multiple points increased by 11.75 and 8.07 percentage points,respectively compared with accurate algorithm and genetic algorithm.This method provides a support for improving the delivery efficiency of cigarette products.