基于气象与非气象因素的客流量单数逐日预测模型

Daily Prediction Model of Passenger Flow Based on Meteorological and Non-meteorological Factors

客流量单数指的是到达商场的顾客真实买单的数量,客单数越多,经营者的工作量和相关消耗就越多,相应的收入也会越多。为建立对超市经营团体乃至整个商业服务行业具有应用价值的定量的短期超市客单数预测模型,选用6种机器学习预测方法进行尝试。结果显示:(1)客单数与气象因子之间的确存在着一定的相关性。客单数会随着气温和舒适度指数的升高,以及风速、相对湿度和降水量的降低,而有所增加。(2)6种机器学习预测方法中,当输入因子选择为气温、风速、相对湿度、降水量级别、舒适度指数、星期、是否节假日、是否节气共8个全因子进行模型训练后,得到的预测效果最佳。相对而言,随机森林的预测效果最优。(3)机器学习方法可以有效地进行客单数回归预测,定量化的预测模型可以作为经营者商业准备行为的科学借鉴,使经营者充分利用好人力和物力成本,有助于科学地节能减排。

Passenger flow refers to the customers who actually pay the bill when they are shopping in markets.The more the customers pay bills,the more workload and related consumption of operators are,and accordingly the more income they obtain.Therefore,a quantitative short-term passenger flow prediction model is very important for market management groups and even for the whole service business.To establish such a model,this paper selects six machine learning prediction methods to give it a shot.The results show that:(1)There is a certain correlation between the passenger flow and meteorological factors.The passenger flow will increase with the increase of temperature and comfort index,and the decrease of wind speed,relative humidity and rainfall.(2)Among the six prediction methods,when the input factors are the 8 factors,namely,air temperature,wind speed,relative humidity,rainfall magnitude,ET index,week,holiday and solar term,the model training can get the best prediction effect.Comparatively,the prediction effect of random forest is the best.(3)Machine learning method can effectively make regression prediction of the number of passengers.The quantitative prediction model can be used as a scientific reference for business preparation behavior of operators,so that the operators can make full use of human and material costs and contribute to energy saving and emission reduction.