Adaptable User Profiles for Intelligent Geospatial Queries

The geospatial information user community is becoming increasingly diverse, with numerous users accessing distributed datasets for various types of applications. Currently in GIS, unlike traditional databases, there is a lack of machine learning algorithms to customize information retrieval results. Thus the particular interests of individual users are not taken into account in traditional geospatial queries. In this paper we present a system that adjusts query results based on user requirements and needs. It does so by using a collection of fuzzy functions that express user preference specifically in GIS environments. The focus of this work is on preference learning for one‐dimensional, quantitative attributes, and on the customization of geospatial queries using this information. The model used to express user preferences adjusts gradually to the underlying complexity during a training process, starting with fairly simple linear functions and progressing to complex non‐linear ones as needed. Our advanced modeling capabilities are demonstrated through an applicability example, and statistical simulations show the robustness of our system.