Knowledge from data-the fuzzy data browser in Fril

When considering large bodies of data, humans generally prefer to work with heuristic (usually fuzzy) rules which summarise patterns in the data. Forming these rules is often a matter of intuition which may be complicated by missing, noisy, or incorrect data; however, a set of fuzzy rules is a highly compressed summary which can be used to predict or verify the data, and is easily understood by a human. This demonstration shows how Fril can model uncertain and incomplete databases, and generate and test hierarchical rules which summarise the data. Fuzzy sets are created automatically, and the importance of different features is determined using semantic unification. Human expertise can be input at any stage, and different rules can be tested against the known cases in the database. We focus on some simple examples to illustrate the use of the fuzzy data browser