Neural network model based on travel planning for travel time prediction

Abstract

The development of smart cities presents new challenge to more convenient and intelligent transportation, while the advent of big data era promotes the sharing of travel data, which provides powerful data support for the development of smart travel. As an important part of smart travel, accurate travel time prediction is definitely crucial. Different from traditional passive forecasting methods which is only based on historical data, this paper proposes a Conv-LSTM network model based on travel planning data for travel time prediction from the perspective of user travel. This model can actively predict the upcoming traffic state according to the upcoming data generated by the user’s travel planning information released before travel. Specifically, we first introduce the definition of travel plan and how to calculate the future planned flow according to travel planning information. Then, the future planned data and the corresponding segment historical data are introduced into the designed Conv-LSTM model to extract spatial-temporal features and then realize the prediction of road travel time. In this study, specific travel path is taken as the research object. Extensive experimental results demonstrate that this method has high accuracy, and remarkably outperforms benchmark methods in various metrics.