基于WTD-CEEMDAN-Bi-LSTM-GRU的共享单车需求预测

Demand Forecast for Shared Bicycles Based on WTD-CEEMDAN-Bi-LSTM-GRU

对共享单车的需求量预测直接影响到共享单车运营企业的调度运营水平与质量。提出一种基于小波阈值降噪(Wavelet Threshold Denoising,WTD),完全自适应噪声集合经验模态分解(Complete EEMD with Adaptive Noise,CEEMDAN),双向长短时记忆网络(Bi-direction Long Short-Term Memory,Bi-LSTM)与门控循环单元网络(Gated Recurrent Unit,GRU)的组合预测模型。选取上海市共享单车出行数据,通过降噪、分解、高低频识别与重构来降低数据的复杂度,以提高预测精度;并通过与不同基准模型的比较,验证所提出的组合预测模型的有效性。结果表明:所提出模型与其他基准模型相比,其预测效果的MAPE、RMSE、R^(2)参数均表现较好,模型具有较强的竞争力。

The demand forecast of shared bicycles directly affects the level and quality of dispatch operation of shared bikes operating enterprises.This paper proposes a combined forecasting model on a basis of wavelet threshold denoising(WTD),complete EEMD with adaptive noise(CEEMDAN),Bi-direction long short-term memory(Bi-LSTM)and gated recurrent unit network(GRU).The travel data of shared bicycles in Shanghai are selected to reduce the complexity of the data through noise reduction,decomposition,high and low frequency identification and reconstruction,so as to improve the prediction accuracy.The validity of the combined prediction model is verified by comparing it with different benchmark models.The results show that compared with other benchmark models,the proposed model has better performance of MAPE,RMSE and R2 parameters,and the model has strong competitiveness.