Demand forecasting method of shared bikes combined with multivariate meteorological factors
In the field of urban transportation,shared transportation has been widely used.As a major mode of transportation,shared bicycles are famous for their efficient mobility and timeliness.However,due to missing observations and randomly changing context conditions in the data set,false correlation between data and features occurs,which makes the prediction of the model fail in some special scenarios.To solve this problem,we use the CNN-BiLSTM-Attention model to predict and analyze the demand for shared bicycles.This study selected bike-sharing data in New York City and focused on analyzing the influence of meteorological factors and time factors on the demand for bike-sharing.The results of data analysis and visualization show that humidity,peak hours and temperature have a significant impact on the demand for bike-sharing.By using the CNN-BiLSTM-Attention neural network model to single-step predict the hourly bikesharing demand,this study selects a variety of mainstream models including LightGBM and Bagging as benchmarks for comparison.The experimental results show that the CNN-BiLSTM-Attention model performs well in the prediction task,its 2 score is as high as 0.952,which is significantly better than other comparison models,and the Root Mean Square Error(RMSE)is 0.018 3.Compared with the best performing baseline model,the RMSE of our model is reduced by 5%.This paper provides data support and decision-making reference for the operators of shared bicycles to formulate scientific management and delivery strategies.
万方数据
维普
