基于多因素稀疏回归预测模型的商家客流量预测  

Retail consumer traffic forecasting based on multi-factor sparse regression prediction model

作　　者：郑增威[1] 杜俊杰 周燕真 孙霖[1] 霍梅梅[1] Zheng Zengwei;Du Junjie;Zhou Yanzhen;Sun Lin;Huo Meimei(Intelligent Plant Factory of Zhejiang Province Engineering Laboratory,Zhejiang University City College,Hangzhou 310015,China;College of Computer Science&Technology,Zhejiang University,Hangzhou 310012,China)

机构地区：[1]浙江大学城市学院智能植物工厂浙江省工程实验室,杭州310015 [2]浙江大学计算机科学与技术学院,杭州310012

出　　处：《计算机应用研究》2020年第5期1440-1444,共5页Application Research of Computers

基　　金：浙江省自然科学基金资助项目(LY17F020008)。

摘　　要：针对智能商业平台中的大数据预测问题,提出一种多因素稀疏回归预测模型。以离散余弦变换为基础,构建包含多个外部因素(节假日、天气、温度)的字典集,通过LASSO方法定量求解稀疏编码模型中各外部因素的影响。实验对2000个商家的客流量进行预测。实验结果表明,外部因素不同程度地影响客流量,在预测模型中叠加外部因素后可以有效提高预测的准确性。同时,与其他方法对比表明,多因素稀疏回归预测模型比RNN、ARIMA等模型的预测效果更好。This paper proposed a multi-factor sparse regression prediction model aiming to solve the problem of big data prediction in business intelligent platform.It constructed a dictionary containing external factors(holidays,weather,and temperature)based on the discrete cosine transform,and quantitatively solved the influence of external factors in the sparse coding model by LASSO.In experiments,it predicted the customer traffics of 2000 stores.The experimental results show that the impact of external factors on the store customer traffic are different,and the prediction accuracy can be effectively improved with the combination of external factors in the prediction model.In addition,compared with other forecasting methods,the result shows that multifactor sparse regression prediction model outperforms than other models such as RNN and ARIMA.

关 键 词：智能商业平台 客流量预测 稀疏回归 多因素分析 字典学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]