O. Farrington, N.K.Taylor
School of Mathematical and Computer Sciences, Dept. of Computing Science, Heriot-Watt University, Edinburgh, Scotland, UK

'Decision tree learning' is one of the most versatile and practical methods of inductive inference developed in applied artificial intelligence. The 'Machine Learning' technique selected for this study was the ID3 algorithm, and represents a new application for this simple classification building technique.
The data, to which the ID3 algorithm was applied, consists of a collection of data sets relating to specific soil analytical procedures (e.g. heavy mineralogy, bulk geochemistry, fabric analysis, etc.) originally collected to address the dark earth conundrum.
In a previous study, PCO analysis proved adept at identifying contributions from individual variables such that it was possible to identify the geo-archaeological components that together constituted the 'character profile' of a typical dark earth.
In this study the classification building abilities of the ID3 algorithms, have been used to explore the possibility of establishing a classification (based on soil data) of not only dark earth's but other archaeological deposits characteristic of urban contexts.
Key words: Machine Learning, ID3 inductive tree, classification, geo-archaeological data, dark earth