A Multifeature Fusion Short-Term Traffic Flow Prediction Model Based on Deep Learnings

Short-term traffic flow prediction is an important component of intelligent transportation systems, which can support trip planning and management. Although existing predicting methods have been applied in the field prediction, they cannot capture complex multifeatures flows resulting unsatisfactory short-term results. In this paper, a multifeature fusion model based on deep learning proposed, consists three modules, namely, CNN-Bidirectional GRU module with attention mechanism (CNN-BiGRU-attention) two Bidirectional modules (BiGRU-attention). The CNN-BiGRU-attention used to extract local trend features long-term dependent flow, BiGRU-attention are daily weekly periodic flow. Moreover, feature layer fuse extracted by each module. And then, number neurons model, loss function, other parameters such as optimization algorithm discussed set up through simulation experiments. Finally, trained tested training test sets from data collected field. results indicate that proposed better achieve has good robustness. Furthermore, compared analyzed against baseline models same dataset, experimental show superior predictive performance models.