Recorded presentation2.6
Genetic Algorithm optimization of principal component approach to clustering for storm type identification
Luna M. Rodriguez, The Pennsylvania State University, University Park, PA
The aim of this work is to develop a unique multi-category classifier to identify storm type based on radar-derived attributes. Due to the quantity of attributes a principal components analysis (PCA) technique is used to reduce the multidimensional dataset to a lower dimension for analysis. The analysis consists of classifying the data into different groups; it is partitioned into subsets (clusters) where they share some common traits. Clustering can distinguish values of some features or characteristics of a situation that might have been overlooked otherwise which then can possibly lead to stratification of data. To optimize this procedure a genetic algorithm (GA) is implemented as a search technique to find exact and approximate solutions to the storm identification. Recorded presentation
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