基于地理加权回归模型的共享电动汽车租赁需求影响因素分析

Analysis of Factors Influencing the Demand for Shared Electric Car Rentals Based on a Geographically Weighted Regression Model

近年来共享经济蓬勃发展,共享电动汽车比传统共享燃油车更加契合环保经济理念,因此得到了快速普及。共享汽车企业以电动汽车作为租赁工具,有助于企业降低成本,迎合市场趋势,并提供更好的用户体验。文章以南京市六个主城区的共享电动汽车数据为例,首先通过对城市多源数据的分析,选取居住用地、工作用地、休闲购物、科教服务、交通设施五个方面作为共享电动汽车设置分配的影响因素,然后对各影响因素按照不同属性进行加权处理,构建了地理加权回归模型,并且与传统的OLS回归模型进行了拟合效果的比较,最后采用空间可视化分析方法分析了共享电动汽车租赁需求量与关键影响因素之间的复杂关系。结果表明:GWR模型能更好地反映共享电动汽车租赁需求量与影响因素之间的关系,具有较高的准确性和更广泛的适用性;工作用地、居住用地、休闲娱乐、交通设施因素在工作日的不同时段对共享电动汽车的租赁需求影响均为促进作用,而在非工作日时这些影响因素对共享电动汽车的租赁需求量具有空间异质性。研究通过使用变量分类加权的方法,提高了预测模型的准确性和实用性,可以为共享电动汽车投放规模确定和日常运营调度提供支持。

In recent years,the flourishing development of the sharing economy has seen a rapid proliferation of electric shared vehicles,which align with environmentally friendly economic principles more than traditional fuel-powered shared cars.Utilizing electric vehicles as rental tools allows sharing economy enterprises to reduce costs,cater to market trends,and provide an enhanced user experience.This article takes the shared electric vehicle data from six main urban areas in Nanjing City as an example.Initially,through the analysis of multi-source urban data,five factors—residential land use,workplace land use,recreational shopping,scientific and educational services,and transportation facilities—are selected as influential factors affecting the allocation and distribution of shared electric vehicles.Subsequently,the variables are weighted differently based on their attributes.A Geographic Weighted Regression(GWR)model is constructed and compared with the traditional Ordinary Least Squares(OLS)regression model to assess the fitting effects.Finally,spatial visualization analysis methods are employed to examine the intricate relationship between the demand for shared electric vehicle rentals and key influencing factors.The results indicate that the GWR model better reflects the relationship between the demand for shared electric vehicle rentals and influencing factors,demonstrating higher accuracy and wider applicability.Factors such as workplace land use,residential land use,recreational activities,and transportation facilities exhibit a promoting effect on shared electric vehicle rental demand during various time segments on workdays.However,on non-workdays,these influencing factors demonstrate both positive and negative effects on the demand for shared electric vehicle rentals.By employing a variable classification weighting method,this study enhances the accuracy and practicality of the predictive model,providing valuable insights for determining the scale of shared electric vehicle deployment and daily operational scheduling.