Use of Data Mining Techniques to Identify Crisis in Dryland Living

Saleh A. Wasimi

Faculty of Business and Informatics, Central Queensland University, Rockhampton, Australia
e-mail: s.wasimi@cqu.edu.au

Abstract

People living in drylands usually adopt their lifestyle to the prevailing harsh environment around them.  Their rudimentary living conditions generally reflect that human lifestyle and environmental factors are intertwined.  People in the arid region are used to some variability in their environment as is to be expected from people who live close to nature.  However, when this variability extends to the level so as to cause stress to the population, it can be termed ‘crisis.’  The objective of this study has been to identify the critical values in hydrological, meteorological, environmental, and socio-economic factors that can trigger the onset of crisis.

The study area is the interior drylands of New South Wales in Australia.  The region has suffered a number of severe droughts in the last three decades.  The amount of data available from direct measurements, remote sensing, secondary sources, and questionnaire survey is phenomenal. Since the volume of data to be dealt with is enormous, and a large part of it is unstructured in the form of textual data, application of data mining tools have been adopted in this study. The standard data mining strategies of classification, clustering, text mining, and association rule mining have been applied.

Essentially, data mining is a black box approach that cannot be conceptualized, but it has found many applications where the problem is too complex or overwhelming with myriad of information.  The findings of this study do not unveil any causal relationships as is typical with data mining techniques.  Nevertheless, they provide some association of conditions that can forewarn an impending crisis for planning mitigating measures.