PROJECT TITLE :

Road Traffic Speed Prediction: A Probabilistic Model Fusing Multi-Source Data - 2018

ABSTRACT:

Road traffic speed prediction could be a difficult downside in intelligent transportation system (ITS) and has gained increasing attentions. Existing works are mainly primarily based on raw speed sensing knowledge obtained from infrastructure sensors or probe vehicles, that, however, are limited by expensive value of sensor deployment and maintenance. With sparse speed observations, ancient strategies primarily based solely on speed sensing information are insufficient, particularly when emergencies like traffic accidents occur. To address the difficulty, this Project aims to improve the road traffic speed prediction by fusing ancient speed sensing knowledge with new-kind “sensing” data from cross domain sources, like tweet sensors from social media and trajectory sensors from map and traffic service platforms. Jointly modeling data from totally different datasets brings many challenges, including location uncertainty of low-resolution information, language ambiguity of traffic description in texts, and heterogeneity of cross-domain data. In response to those challenges, we gift a unified probabilistic framework, referred to as Topic-Enhanced Gaussian Method Aggregation Model (TEGPAM), consisting of three components, i.e., location disaggregation model, traffic topic model, and traffic speed Gaussian Method model, that integrate new-type knowledge with traditional information. Experiments on planet knowledge from two massive cities validate the effectiveness and efficiency of our model.