Date: 19 Oct 2006

Knowledge discovery in spatial data by means of ILP

* Final gross prices may vary according to local VAT.

Get Access

Abstract

We show that inductive logic programming (ILP) is a powerful tool for spatial data mining. We further develop the direction started (or symbolised) by GeoMiner [9] and argue that the technique developed for database schema design in deductive object-oriented databases is fully usable for spatial mining and overcome, in expressive power, some other mining methods. An inductive query language, with richer semantics, is proposed and three kinds of inductive queries are described. Two of them are improved versions of DBMiner [8] rules. The third kind of rules, dependency rules, allow to compare two or more subsets. Then a description of GW iM mining system as well as results reached by the system are given. We conclude with discussion of weaknesses of the method.