Deep Ensemble Learning Model for Long-Term Travel Time Prediction on Highways

Travel time plays a vital role in people's daily lives. It can help them not merely avoid traffic congestion but save time as well. Additionally, on account of the interactive influence of time, we divide data into long and short sequences and exploit data to predict next-hour travel time. We introduce a deep ensemble learning model which predicts the traveling time by taking tendency, deriving by XGBoostm and recency embeddings, deriving by a deep neural network, into account. The experiments on the real dataset shows that the proposed model outperforms other state-of-the-art models.