Knowledge discovery from area-class resource maps: Data preprocessing for noise reduction

Research output: Contribution to journalArticlepeer-review

24 Scopus citations

Abstract

Spatial data mining techniques have been investigated to extract valuable knowledge from natural resource maps. One challenge in applying general-purpose data mining algorithms to knowledge discovery from these maps is that the unavoidable existence of errors in traditional area-class resource maps introduces noise to the process and impacts the accuracy of the extracted knowledge. This paper examines the effect of data preprocessing in knowledge discovery from area-class resource maps. Presented here is a sampling method designed to obtain samples that exclude noise and are representative of the central concepts of the mapped classes. The method is based on selecting only the pixels that fall in modes of environmental histograms for each individual class. A case study in extracting knowledge from soil maps shows that data preprocessing plays an important role in the knowledge discovery process, and that the sampling method based on environmental histograms significantly improves the knowledge discovery performance.

Original languageEnglish
Pages (from-to)297-308
Number of pages12
JournalTransactions in GIS
Volume8
Issue number3
DOIs
StatePublished - Jun 2004

Fingerprint

Dive into the research topics of 'Knowledge discovery from area-class resource maps: Data preprocessing for noise reduction'. Together they form a unique fingerprint.

Cite this