Visual Data Mining in Spatial Interaction Analysis with Self-Organizing Maps

Abstract

Given that many spatial interaction (SI) systems are often constituted in large databases with high thematic dimensionality, data complexity reduction tasks are essential. The opportunity exists for researchers to examine the formation of different types of SIs as well as their interdependencies by exploring the patterns embedded in the data. To circumvent the limitations of existing methods of flow data compression and visual exploration, we propose an integrated computational and visual approach, known as VISIDAMIN, for handling both SI data projection and SI data quantization at once. The computational method of self-organizing maps serves as the data mining engine in this process. Using a large domestic air travel dataset as a case study, we examine how the characteristics of the air transport system interact with the SI system to create relationships and structures within the US domestic airline market.

The linkage between the lifestyle of knowledge-workers and their intra-metropolitan residential choice: A clustering approach based on self-organizing maps

“…Skupin and Hagelman (2005) showed the usefulness of SOM in exploring temporal demographic trends across spatial data units (e.g., counties) over time. Yan and Thill (2009) used SOM to perform both cross-sectio nal and temporal analysis of the airline market in the US Arribas-Bel, Nijkamp, and Scholten (2011) employed SOM to explore the key dimensions of urban sprawl patterns in Europe.…”

The Creative Urban Diaspora Economy: A Disparity Analysis Among Migrant Entrepreneurs

“…It is a data-reduction technique that was first developed in relation to the study of the spatial organization of brain functions and aimed to perform two main types of compression: on the one hand, it shrinks the number of input observations (quantization ); on the other hand, it compresses the number of dimensions or attributes of each observation, usually to two of them (projection ). Although it has not been widely applied in the context of social sciences until recently (see e.g., Skupin and Hagelman 2005, Spielman and Thill 2008, Yan and Thill 2009, ArribasBel et al 2011, the SOM features a number of characteristics that make it very useful for exploration and presentation of complex relationships buried in high-dimensional socioeconomic datasets. This in turn converts it in a useful device to present information in an intuitive way to a non-technical audience, for instance, in the context of decision making support.…”

High Performance in Complex Spatial Systems: A Self-Organizing Mapping Approach with Reference to the Netherlands

“…It is in this context that the term 'data mining' fits, and where the SOM, as one of the components of this new toolkit, gains the power to shed new light on, and uncover, interesting patterns in complex relations. Recent attempts to explore this methodology in the social sciences include studies by: Skupin and Hagelman (2005) and Spielman and Thill (2008) in sociodemographics; Yan and Thill (2009) The output of a SOM is a network of neurons that are interconnected by topological relationships, and that are usually represented by hexagons, implying therefore that the normal neuron (i.e. not on the edge) will have six neighbours and will thus be connected to six other neurons.…”

“…In this context, dissimilarity between observations is translated into distance within a SOM: similar observations will tend to be close to each other, whereas distinct ones will be more distant. Although the SOM has been compared with other statistical techniques, such as principal components and factor analysis, or clustering algorithms, such as k-means or hierarchical clustering, according to Yan and Thill (2009) there are clear advantages inherent in this approach: the learning nature of the algorithm which avoids no recovery, continuing to compare and use information from every observation even after they are first assigned to a neuron; the use of the distance of each input from all neurons as opposed to only the nearest one; the fact that the SOM is a combination of both data quantization and data projection; and the visualization opportunities that it offers. Among all of these, the latter one has greater implications for the present paper, since it is the ability to represent in a visual and intuitive way the underlying statistical properties of high-dimensional data sets that makes the SOM a suitable tool to unfold complexity into understandable patterns that enhance understanding.…”