摘要
卷烟的销量预测可以为卷烟的运输、配送、投放提供指导,使烟草行业能更好适应市场的变化需求。卷烟销量受多方面的因素影响,具有地域性、季节性和周期性等规律,为科学制定市级烟草公司的卷烟投放策略,提出采用时间序列模型和xgboost机器学习模型构成的混合算法模型对单品销售量前十名的卷烟进行预测。通过xgboost机器学习模型对波动较大的销售数据进行处理,使整个销售数据形态趋于相对稳定,通过时间序列模型对不同特征值的销售数据进行预测,形成最终预测成果,以山东省某地市公司的卷烟销售数据为例,对所提方法进行验证,验证结果表明,该模型具有较高的预测精度,能够准确的反映卷烟销量的变化趋势。对比实验也表明,本文所提出的方法比其他几种方法预测精度高,可以为卷烟销售提供一定的科学依据。
Cigarette sales forecast can provide guidance for the transportation,distribution and distribution of cigarettes,so that the tobacco industry can better adapt to the changing market needs.Cigarette sales are influenced by many factors,has a regional,seasonal and periodic law of city level tobacco monopoly bureau for science of cigarettes on the strategy,put forward using xgboost machine learning model and time series model of hybrid algorithm model item sales to dealers for the unit of the top 10 cigarettes.Through xgboost machine learning model to deal with volatile sales data,make whole sales data forms tend to be relatively stable,through the time series model to forecast the different characteristic value of sales data,form the final prediction results,in shandong province city of cigarette sales data,for example,to validate the proposed method,the verification results show that the model has higher prediction accuracy,can accurately reflect the change trend of tobacco sales.The comparison experiment also shows that the prediction accuracy of the proposed method is higher than that of other methods,which can provide scientific basis for cigarette sales.
作者
宋楠
刘际鑫
李林
仇道霞
SONG Nan;LIU JI-xin;LI Lin;QIU Dao-xia(Shandong Tobacco Monopoly Bureau,Shan dong Jinan 250101,China)
出处
《新一代信息技术》
2021年第13期17-25,共9页
New Generation of Information Technology
关键词
卷烟
销售预测
时间序列模型
机器学习
Cigarettes
Sales forecast
Time series model
Machine learning
