Time series forecast of sales volume based on XGBoost

作     者：Lingyu Zhang Wenjie Bian Wenyi Qu Liheng Tuo Yunhai Wang 

作者机构：School of Computer Science and Technology Shandong University Qingdao China Didi AI Labs Didi Chuxing Beijing China Smart Transportation Center National Engineering Laboratory for Big Data Analysis and Applications Beijing China 

出 版 物：《Journal of Physics: Conference Series》 

年 卷 期：2021年第1873卷第1期

学科分类：07[理学] 0702[理学-物理学] 

摘      要：Some problems such as the decline of new labor force, the increase of retired labor force emerge because of the complex and changeable market environment, consequently exacerbating the staffing problem in the retail industry. Also, the unreasonable distribution of personnel, there are few people in busy hours and too many people in idle hours, which causes waste of labor. To address this issue, we analyzed the time series of sales volume in the retail industry in detail, and processed the data with feature engineering for predicting the in-store sales volume in the future. At the same time, other features such as weather and temperature are added to improve the accuracy of the model. Considering the characteristics of the data, we choose XGBoost as the prediction model. The experiments on real-world datasets verified better performance of proposed model compared with other state-of-art models.