TY - JOUR

T1 - Acquiring transition rules between multiple representations in a GIS

T2 - An experiment with area aggregation

AU - Sester, Monika

PY - 1999/1/1

Y1 - 1999/1/1

N2 - Multi-scale representation is an issue of growing interest and importance in geographic information systems. It deals with the representation of spatial entities at different resolutions in one common information system. Such representations have multiple benefits, e.g. the transition from one scale to another and especially the use of coarse-to-fine approaches for data analysis. Multi-scale representations can be derived in two ways. In one way, the data of different scales are acquired separately and the links between the scales are established subsequently. The other possibility is to derive the series of representations from a single, most detailed representation. This involves the availability of procedures defining the possible transitions objects undergo when moving from one scale to the next, i.e. database generalization procedures. For small changes in scale, simple smoothing operations can be applied. At a certain level, however, there are gaps in the representation which cannot be reflected by elementary processes, but have to be represented by symbolic descriptions, e.g. a set of rules. Such rules may be known in advance and directly programmed into a system. Often, however, knowledge is not available in an explicit form. The idea of this contribution is to use the machine learning technique 'learning from examples' to derive these rules. The examples are taken from existing data sets-the system automatically derives the transition rules from them.

AB - Multi-scale representation is an issue of growing interest and importance in geographic information systems. It deals with the representation of spatial entities at different resolutions in one common information system. Such representations have multiple benefits, e.g. the transition from one scale to another and especially the use of coarse-to-fine approaches for data analysis. Multi-scale representations can be derived in two ways. In one way, the data of different scales are acquired separately and the links between the scales are established subsequently. The other possibility is to derive the series of representations from a single, most detailed representation. This involves the availability of procedures defining the possible transitions objects undergo when moving from one scale to the next, i.e. database generalization procedures. For small changes in scale, simple smoothing operations can be applied. At a certain level, however, there are gaps in the representation which cannot be reflected by elementary processes, but have to be represented by symbolic descriptions, e.g. a set of rules. Such rules may be known in advance and directly programmed into a system. Often, however, knowledge is not available in an explicit form. The idea of this contribution is to use the machine learning technique 'learning from examples' to derive these rules. The examples are taken from existing data sets-the system automatically derives the transition rules from them.

KW - Database generalization

KW - Machine learning

KW - Model-based generalization

KW - Multi-scale representation

UR - http://www.scopus.com/inward/record.url?scp=0032978982&partnerID=8YFLogxK

U2 - 10.1016/S0198-9715(99)00006-X

DO - 10.1016/S0198-9715(99)00006-X

M3 - Article

AN - SCOPUS:0032978982

VL - 23

SP - 5

EP - 17

JO - Computers, Environment and Urban Systems

JF - Computers, Environment and Urban Systems

SN - 0198-9715

IS - 1

ER -