基于GS⁃XGBoost 的共享单车需求预测分析研究

Research on forecasting and analysis of shared bicycle demand based on GS⁃XGBoost

随着共享单车的大量涌入,为人们带来方便的同时,也存在单车分布不均衡、用户借车困难、体验差等问题。为准确预测城市各个站点每小时共享单车的需求量,解决各个站点之间供需不平衡问题,引入了一种网格搜索优化XGBoost的预测模型,即GS⁃XGBoost。研究共享单车需求影响因素并利用Pearson相关性分析法分析特征的影响因素,提取特征值,构建输入序列进行模型训练,并且与传统的模型进行对比分析。结果表明,GS⁃XGBoost模型能够很好地预测共享单车的需求量,有着更低的MSE、MAE,以及更高的R2,有助于提高共享单车需求预测的精确度。

With the influx of shared bicycles,while bringing convenience to people,there are also problems such as uneven distribution of bicycles,difficulty in borrowing bicycles,and poor experience.In order to accurately predict the hourly demand for shared bicycles at various sites in the city and solve the problem of unbalanced supply and demand between various sites,a grid search optimization XGBoost prediction model,namely GS‑XGBoost,is introduced to study the factors affecting the demand for shared bicycles.Pearson correlation analysis method was used to analyze the influencing factors of the characteristics and extract the characteristics value,construct the input sequence for model training,and conduct comparative analysis with traditional models.The results show that the GS‑XGBoost model can predict the demand for shared bicycles very well,with lower MSE and MAE and higher R2,helps to improve the accuracy of demand forecasting for shared bikes.