Data mining and soil salinity analysis

Abstract

The paper explores connections between decision support systems, remotely sensed and GIS data for environmental planning, and monitoring secondary soil salinization. The paper introduces a decision support knowledge base system called SALT MANAGER used as the starting point for three experiments in data mining which structure the paper. In the first, classified GIS data is passed to an inductive learning programme. The task is to reconstruct the classification rules. The resulting rules are used to improve knowledge base system performance. The second experiment reports on patterns of attribute value combinations occurring for specific classification classes. These patterns can be used to elicit new knowledge in the domain and lead to a form of knowledge discovery. This process is commonly referred to as ‘data mining'. The third experiment measures the effect of an additional electromagnetic data layer on the knowledge base system. Again, our efforts yield novel knowledge discovery results from the application of data mining techniques. Finally, a comparison of different machine learning algorithms in the secondary salinization domain is given.