基于长短期记忆神经网络模型的共享单车短时需求量预测

共享单车具有很强的流动性和高随机性,为了更加准确地预测某区域内每小时的单车使用数量,通过爬取纽约市Citi Bike共享单车的天气特征数据信息,并分析时间因子、气象因子等对单车需求量的影响;采用长短期记忆(long short-term memory,LSTM)神经网络模型预测共享单车的短期需求量,并与传统的循环神经网络(recurrent neural network,RNN)和BP(back-propagation)神经网络模型预测结果进行比较。实验结果表明:影响单车需求量的主要因素包括温度、节假日、季节以及早晚高峰时间段等因素;与传统BP神经网络算法和循环神经网络RNN算法相比,LSTM鲁棒性高,泛化能力强,且预测结果曲线与真实结果曲线相吻合;预测精度高(精确度为0.860)均方根误差最小(为0.090),误差小。可见LSTM模型可以用来对共享单车的短时需求量进行预测。

Short-term Demand Forecasting of Shared Bicycles Based on Long Short-term Memory Neural Network Model

Shared bicycles have strong liquidity and high randomness. In order to more accurately predict the number of bicycles per hour in an area, by crawling the weather characteristics data of bicycles shared by Citi Bike in New York City, and analyzing the influence of time factor and meteorological factors on the demand for bicycles; the Long Short-Term Memory (LSTM) neural network model was used to predict the short-term demand of shared bicycles for shared bicycles, and compared with the traditional Recurrent Neural Network(RNN) and Back-Propagation Network(BP) neural network model prediction results. The experimental results show that the main factors affecting the demand for bicycles including temperature, holidays, seasons and morning and evening peak time periods. Compared with traditional BP neural network algorithm and cyclic neural network RNN algorithm, LSTM has high robustness and strong generalization ability; and the prediction result curve is consistent with the real result curve; the prediction accuracy is high (Acurracy=0.860) and the root mean square error is the smallest (RMSE=0.090), and the error is small; It can be seen that the LSTM model can be used to predict the short-term demand for shared bicycles.