Public Bicycle Demand Forecast Based On PCA-BA-GRNN Model

In recent years,with the development of economy and the improvement of people's lives,the number of motor vehicles has continued to rise.The automobile brings people comfort and convenience,but it also brings a lot of problems: traffic jams,environmental pollution,and energy consumption.As a convenient and efficient,green transportation,bicycles are becoming more and more popular.Various cities began to build their own public bicycle systems to improve the citizens' short-distance travel efficiency.However,each bicycle company is faced with the problem of balancing the supply and demand of bicycles,and there are many factors that affect bicycle demand,such as season,time,temperature,wind speed,humidity and holidays.So a precise and effective bicycle demand forecasting model can not only effectively manage and rationally schedule bicycles,but also maximize the utilization rate of bicycle resources and reduce management costs.Generalized Regression Neural Network(GRNN)has been widely applied to nonlinear problems with high fault tolerance,good nonlinear mapping ability and stability.Especially for small sample data,the GRNN model has fast convergence and generalization ability.The parameters in the GRNN model affect the prediction results.Optimizing this parameter by the intelligent algorithm can improve the prediction ability of GRNN.At the same time,the noise properties and the attributes that have a weak influence on the results are removed through dimension reduction methods,so that the training efficiency of the neural network is improved and the performance of the prediction is improved.Therefore,this paper proposes a PCA-BA-GRNN model to predict the demand of urban public bicycles.The work of this article is as follows:(1)In this paper,several popular prediction models are studied.The structure,theoretical basis and advantages of GRNN are elaborated in detail.And we pointed out that the generalized performance of GRNN can be improved by the intelligent optimization algorithm.(2)This paper proposes a PCABA-GRNN model,which can improve the forecasting performance of bicycle demand by the principal component analysis of the bicycle demand attributes,and then using the bat algorithm to optimize the GRNN model.(3)Bicycle demand data visualization analysis.Describe in detail the characteristics of the bicycle data set and various factors and laws that affect the bicycle demand;(4)Verify the predictive power of the model.Experiments show that the PCA-BA-GRNN model presented in this paper has better prediction effect than the decision tree,support vector machine,and Bagging model.The bat algorithm optimizes the parameters of the GRNN model to improve its prediction accuracy,and the principal component analysis dimension reduction can shorten the training time of the model,reduce the interference of irrelevant variables on the training,and maximize the prediction accuracy of the model.