Environmental modelling with machine learning

[2014-09-12 Fri 19:00] speaker: Shamraz, Department of Geography event: Friday Hacks, NUS Hackers ** AI is whatever that hasn’t been done yet ** How much water is flowing through the river? ** Missing data equipment failure out of power ** Use machine learning to fill in the missing data ** Process based models vs Data driven models Use knowledge of hydrology to model the physical processes vs Use collected data to infer the model from the data ** Classify regions on a map into the category it belongs to ** Predicting streamflow from rainfall and temperature *** Linear regression streamflow = A x rainfall + B x temperature + C *** Decision tree *** Neural network *** SVM ** Additional additional variables streamflow before, after rainfall before, after, average ** Different variations proximal training - use only recent data instead of all data model seasons separately as the data is different across seasons