Evaluation of urban bus service reliability on variable time horizons using a hybrid deep learning method

•A LSTM deep learning method for transit arrival prediction is evaluated.•The VMD is adopted to model average bus link speed series into several sub-layers.•The LSTM network is adopted as the predictor of each sub-layer.•The VMD-LSTM model provides satisfactory bus speed forecasting results.•The met...

Ausführliche Beschreibung