Extended Abstract (40K)2.4
Using Multiple Machine Learning Techniques to Improve the Classification of a Storm Set
David John Gagne II, University of Oklahoma School of Meteorology, Norman, OK; and A. McGovern
Our approach to the AI contest is derived from our research in classifying simulated and observed storms using decision trees. Based on our results in that project, we transformed the data for the contest in several non-linear ways and developed a multi-algorithm approach that outperforms the baseline decision tree provided by the contest. Our approach uses a combination of decision trees, neural networks, and boosting to generate the final model.
Session 2, Forecasting contest submissions (Participants will present the results of their model forecasts of the posted dataset)
Tuesday, 22 January 2008, 1:30 PM-3:15 PM, 205
Previous paper Next paper
Recorded presentation