Attribute selection for environmental science problems

Over the past few years, many data mining techniques have been developed, but encouraging the use of uncommonly known algorithms is equally important. This book represents the PhD thesis written by the author and has three fundamental aims; 1) to develop a flexibly applicable new attribute selection method, Tree Node Selection (TNS), 2) to introduce the use of theoretical or uncommonly used computer algorithms, the K-Maximum Subarray Algorithm and Ant-Miner, and the unique advanced statistical and mathematical method, Singular Spectrum Analysis, as a data pre-processing method for C4. 5 algorithm, and 3) to encourage environmental scientists to use these methods. The methods were tested with benchmark data and various environmental science problems: sea container contamination, the Weed Risk Assessment model and weed spatial analysis for New Zealand Biosecurity, air pollution, climate and health, and defoliation imagery. The analyses and methodologies that are introduced in this book will be useful for students and professionals to investigate many other environmental science problems to help improve the policy and management processes.