面向连锁销售大数据的劳动力分析预测系统设计与实现

发布时间：2018-02-14 17:08

本文关键词： 连锁零售企业劳动力模型劳动力需求预测机器学习 B/S/S UML　出处：《山东大学》2015年硕士论文　论文类型：学位论文

【摘要】：连锁零售企业具有门店多、分布广、员工多等特点,人力成本在企业毛利中所占比重很高。每家零售企业都会设法提高利润率,比如推出促销政策、实行会员制、控制成本等。其中,通过有效管控劳动力需求,可以减少人力资源消耗,降低人力成本,是控制成本的一个有效途径。现在是一个大数据的时代,各行各业都在讨论应用大数据,从数据中掘金,连锁零售企业理应成为大数据应用的主力军。因为连锁零售行业是一个较早进入信息化的行业,历年来产生了大量的营销数据,这些数据一般都做到了很好的记录和保存,这些数据中隐含着规律和提高效益的途径,企业需要利用起这些大数据对未来做出预测。但是于此同时,连锁零售企业的销售数据又具有数据体量大、数据分布广、数据产生速度快等特点,数据波动受促销手段、社会政策、公休假期、恶劣天气甚至重大活动等因素影响明显,这些特点都对大数据预测应用的性能和准确性提出了挑战,加大了预测的难度。针对零售销售大数据对预测带来的挑战和难度,本文设计了零售企业各门店业务大数据的收集方案；建立了用于进行劳动力需求预测的企业劳动力模型；充分考虑零售销售大数据的特点和影响因素,建立了以机器学习算法为核心的科学有效预测机制。本文分别对系统的需求分析、概要设计、详细设计和实现各个阶段详细阐述,并重点描述了应用连锁销售大数据进行预测的关键问题的解决方案。在设计和实现过程中,利用面向对象的分析和设计思想,采用用例图、类图、顺序图、活动图等方法对系统进行UML可视化建模,阐述了系统主要模块和关键业务的设计思路和方法。本文选用JBoss Seam技术框架,采用B/S/S三层体系架构。学习并应用了优秀的数据处理转换工具Kettle,完成了数据归集工作。通过集成应用Weka等提供的基于时间序列的机器学习算法实现了服务指标的科学预测。通过应用JPPF实现了多任务分布式并行预测,通过应用数据库存储过程将数据处理压力向数据库转移,提高了预测性能。本系统已在美国Bigy连锁超市集团部署运行,目前系统运行稳定,用户使用本系统实现了促销商品、非促销商品和服务指标的有效预测,构建了劳动力模型,帮助他们实现了对各门店劳动力需求的有效预测和管理。
[Abstract]:Chain retail enterprises have the characteristics of many stores, wide distribution, many employees, etc. The proportion of labor cost in gross profit is very high. Every retail enterprise will try to improve its profit margin, such as promoting sales promotion policy and implementing membership system. Cost control and so on. Among them, by effectively controlling the demand for labor force, we can reduce the consumption of human resources and reduce the cost of manpower, which is an effective way to control costs. Now it is an era of big data. All kinds of industries are discussing the application of big data, excavating gold from the data, and chain retail enterprises should be the main force in big data's application. Because the chain retail industry is an industry that entered information earlier, it has produced a large amount of marketing data over the years. These data are generally well recorded and preserved. There are hidden rules and ways to improve efficiency in these data. Enterprises need to use these big data to predict the future. But at the same time, The sales data of chain retail enterprises are also characterized by large volume of data, wide distribution of data, fast data generation, and so on. Data fluctuations are obviously affected by such factors as promotion measures, social policies, public holidays, bad weather and even major activities. These characteristics have challenged the performance and accuracy of big data's prediction application, and increased the difficulty of forecasting. This paper has designed the collection scheme of big data, established the labor force model used to forecast the labor demand, fully considered the characteristics and influence factors of the retail sales big data, A scientific and effective prediction mechanism with machine learning algorithm as the core is established in this paper, the requirements analysis, outline design, detailed design and implementation of the system are described in detail. In the process of design and implementation, using object-oriented analysis and design idea, use case diagram, class diagram, sequence diagram, use case diagram, class diagram, sequence diagram, use case diagram, class diagram, sequence diagram, use case diagram, class diagram, sequence diagram, use case diagram, class diagram, sequence diagram, use case diagram, class diagram and sequence diagram. The method of activity diagram is used to visualize the UML modeling of the system, and the design ideas and methods of the main modules and key business of the system are expounded. In this paper, the technical framework of JBoss Seam is selected. Using the three-tier architecture of B / S / S, we have studied and applied the excellent data processing tool Kettle. we have completed the work of data collection. By integrating the machine learning algorithm based on time series provided by Weka and so on, we have realized the service index. Scientific prediction. Multitask distributed parallel prediction is realized by using JPPF. The data processing pressure is transferred to the database by using the database stored procedure, and the prediction performance is improved. The system has been deployed and run in the United States Bigy supermarket chain group. At present, the system is running stably, and the user has realized the promotion of goods by using the system. The effective forecast of non-promotion goods and service index constructs the labor force model and helps them to realize the effective forecast and management of the labor demand of the stores.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP311.52

【相似文献】