Non-parametric regression for space-time forecasting under missing data

► A range of space-time non-parametric regression models are developed for spatio-temporal forecasting under missing data. ► The assumed spatio-temporal autocorrelation is captured by the model. ► The models forecast with reasonable accuracy under the assumption of complete missing data. ► Strong performance is shown in predicting abnormal traffic conditions. ► The model outperforms historical average benchmark model.