基于GTO-CNN-BiLSTM模型的公交车到站时间预测

Prediction of Bus Arrival Time Based on GTO-CNN BiLSTM Model

提升公交车到站时间预测精度可以提高乘客出行效率和公交服务质量、节省公交运营成本。通过分析公交车运行的影响因素、周期与相关性,文章建立了基于人工大猩猩部队算法的卷积双向长短期记忆神经网络(GTO-CNN-BiLSTM),通过人工大猩猩部队算法进行超参数寻优,获得更好的预测效果,采用呼和浩特62路公交到站时间数据进行预测,验证模型预测精度。研究表明:不论是在工作日还是非工作日,早晚高峰还是平峰,GTO-CNN-BiLSTM都能有最优预测效果,相较于卷积双向长短期记忆神经网络(CNN-BiLSTM)、双向长短期记忆神经网络(BiLSTM)和长短期记忆神经网络(LSTM),GTO-CNN-BiLSTM预测结果的平均绝对误差至少减少7.57%,均方根误差至少减少3.84%,平均绝对百分比误差至少减少7.86%。

Accurate prediction of bus arrival time can improve passenger travel efficiency,improve bus service quality,and save bus operating costs.By analyzing the influencing factors,cycles,and correlations of bus operation,this article establishes the Artificial Gorilla Troop Optimizer Convolutional Neural Network Bi-directional Long Short-Term Memory(GTO-CNN-BiLSTM),and uses the GTO for hyperparameter optimization to obtain the better prediction results.The prediction accuracy of the model was verified by using the arrival time data of bus route 62 in Hohhot.The results show that GTO-CNN-BiLSTM has the best predictive performance during morning and evening peak hours or off-peak hours on workdays and non-workdays.Compared to Convolutional Neural Network Bi-directional Long Short-Term Memory(CNN-BiLSTM),Bidirectional Long Short Term Memory Neural Network(BiLSTM),and Long Short Term Memory Neural Network(LSTM),the average absolute error of GTO-CNN-BiLSTM prediction results is reduced by at least 7.57%,the root mean square error is reduced by at least 3.84%,and the average absolute percentage error should be reduced by at least 7.86%.