@inproceedings{dnn-based-prediction-model-spatial-temporal-data, author = {Zhang, Junbo and Zheng, Yu and Qi, Dekang and Li, Ruiyuan and Yi, Xiuwen}, title = {DNN-Based Prediction Model for Spatio-Temporal Data}, booktitle = {}, year = {2016}, month = {October}, abstract = {Advances in location-acquisition and wireless communication technologies have led to wider availability of spatio-temporal (ST) data, which has unique spatial properties (i.e. geographical hierarchy and distance) and temporal properties (i.e. closeness, period and trend). In this paper, we propose a Deep-learning-based prediction model for Spatio-Temporal data (DeepST). We leverage ST domain knowledge to design the architecture of DeepST, which is comprised of two components: spatio-temporal and global. The spatio-temporal component employs the framework of convolutional neural networks to simultaneously model spatial near and distant dependencies, and temporal closeness, period and trend. The global component is used to capture global factors, such as day of the week, weekday or week-end. Using DeepST, we build a real-time crowd flow fore-casting system called UrbanFlow1. Experiment results on diverse ST datasets verify DeepST's ability to capture ST data's spatio-temporal properties, showing the advantages of DeepST beyond four baseline methods.}, publisher = {ACM SIGSPATIAL 2016}, url = {https://www.microsoft.com/en-us/research/publication/dnn-based-prediction-model-spatial-temporal-data/}, address = {}, pages = {}, journal = {}, volume = {}, chapter = {}, isbn = {}, }