Predictive spatio-temporal modelling with neural networks

Hongbin Liu studied the predictive spatio-temporal modelling using Neural Networks. Predictive spatio-temporal modelling is a challenge task due to the complex non-linear spatio-temporal dependencies, data sparsity and uncertainty. Hongbin Liu investigated the modelling difficulties and proposed three novel models to tackle the difficulties for three common spatio-temporal datasets. He also conducted extensive experiments on several real-world datasets for various spatio-temporal prediction tasks, such as travel mode classification, next-location prediction, weather forecasting and meteorological imagery prediction. The results show our proposed models consistently achieve exceptional improvements over state-of-the-art baselines.