Academic journal article Cartography and Geographic Information Science

The Integration of Geographic Visualization with Knowledge Discovery in Databases and Geocomputation

Academic journal article Cartography and Geographic Information Science

The Integration of Geographic Visualization with Knowledge Discovery in Databases and Geocomputation

Article excerpt

Introduction

This paper addresses the use of visualization within exploratory analysis, and data mining and geocomputation, with the overall focus directed to the task of knowledge construction. It represents one component of the broader International Cartographic Association (ICA) Commission on Visualization and Virtual Environments led research agenda effort detailed in this volume (for more on this and related ICA activities, see: www.geovista.psu.edu/icavis/draftAgenda.html).

The major objective of this paper is to investigate the possibilities of applying geographic visualization in each of the problem-solving phases of knowledge discovery, specifically as they relate to geography, and from this to propose a suitable research agenda. The connections between the discovery of generic knowledge and information visualization are described by Card et al. (1998). This paper extends some of those thoughts and ideas into the geographical realm. We necessarily assume the reader has some background knowledge on this topic. We provide citations to introductory material where space does not permit a longer explanation.

This paper is organized as follows: the first section introduces the task of knowledge discovery using visualization and geocomputation tools and motivates interest in this process; the second section defines the state of the art in the joint areas of interest; and the third section distills the research challenges arising from them. These themes are approached from the perspective of scientific inference, specifically the types of inferential mechanisms that knowledge discovery requires; the degree to which they are supported in the various tools and methods described; and the roles played by the system and the user.

The Problem

Geographic datasets continue to become more complex--many areas of geographic analysis now have access to vast digital datasets, and, conversely, many conventional datasets now contain either explicit or implicit spatial references, providing huge and previously untapped data resources. Furthermore, and for the first time, integration of data from different scientific and business communities is becoming a practical reality. To address social and environmental concerns, geographers are linking disparate datasets together across place, scale, time, theme, and discipline and are now beginning to ask new types of questions that were hitherto not possible. The resulting databases are rich in terms of attribute depth and large in the sense of having many records or objects represented. Consequently, uncovering and understanding real-world patterns, or real-world processes that can be found in these data presents a difficult challenge.

Contributing Technologies

Knowledge Discovery

Knowledge discovery has become the focus of major research efforts within the computer science and very large databases (VLDB) communities (as evidenced by many new conference series and journals). The aims of knowledge discovery are very similar to those of visual approaches to exploratory data analysis: i.e., to find useful and valid structure in large volumes of data, and to provide some means of explaining those patterns (Fayyad at al. 1996a). Fayyad et al. (1996b) describe the knowledge construction process as comprising five stages: data selection, pre-processing, transformation, data mining, and interpretation/evaluation. These stages progressively refine a large dataset to the point where it makes sense to propose object structures and relationships.

There is an important distinction between machine-based knowledge and human-cognitive knowledge. The former is a syntactic form of knowledge, while the latter is a semantic form of knowledge. Machine-based knowledge discovery represents observations or assertions (often associated with a degree of confidence), which have been distilled from a dataset. It is thus not "knowledge" in the human-cognitive sense, although it may prompt or compliment human knowledge as part of the visualization process. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.