Classification of changing regions using a temporal signature of local spatial association

Abstract

Spatially associated patterns are often found in geographical phenomena, since nearby entities are often more related than distant ones. Such spatial association also changes over time; hence, the temporal aspect of spatial association needs to be examined using both spatiality and temporality. This paper describes a method of modeling the temporal signatures of spatial association, and thus of grouping similar changes. We employed a Moran scatterplot to assess the local characteristics of a spatial association and then extended it to a time-series Moran scatterplot quadrant signature (MSQS) to capture spatiotemporal changes in regions categorically. We used sequence comparison and data grouping techniques to classify similar regions in terms of the time-series MSQS. We tested the feasibility of the proposed method using a case study of a twenty-four-month (June 2004–May 2006) housing price index for sixty-nine administrative units in the Seoul Metropolitan Area, South Korea.

Probabilistic methods of analysis for the time series Moran scatterplot quadrant signature

“…For comparison, the method of Ahn et al . () was also applied to this dataset. In order to select k * , they recommend the method of Feoli and Lausi ().…”

“…In this simple view, each region is assigned to one of four states or one of four quadrants on the Moran scatter plot, that is, (low, low), (low, high), (high, low) and (high, high) where the first part refers to the region itself and the second refers to its neighbours. Recently, it has been suggested that time can be incorporated into this setting by breaking time into discrete regions and considering a sequence of local Moran statistics over time (Ahn et al ., ). This approach summarises the data as many sequences of discrete variables with four states.…”

“…In order to deal with this problem, a large number of heuristics have been suggested such as the method of Feoli and Lausi () as used by Ahn et al . (). The lack of consensus in determining the number of clusters is evident in the work of Salvador and Chan () who surveyed five methods for selecting the number of clusters.…”

“…Ahn et al . () computed a kernel matrix using a modified Levenshtein metric between regions. The main achievement in the approach was the ability to cluster regions together based on the similarity of a local sequence of Moran statistics, or as the author's refer to it, the Moran scatterplot quadrant signature (MSQS).…”

Complexity reduction in choropleth map animations by autocorrelation weighted generalization of time-series data

“…Autocorrelation on a global and local level is assessed and further processed utilizing the concept of Moran scatterplots. Introduced as a diagnostic tool for visualization of spatial autocorrelation (Anselin, 1993), Moran scatterplots were used for spatial outlier detection (Shekhar, Lu, & Zhang, 2003), to analyze categorical changes of spatial association over time (Ahn, Kim, & Lee, 2009; Rohde, Corcoran, McGee, Wickes, & Townsley, 2015), or to take into account spatial structures within choropleth map classification (Traun & Loidl, 2012).…”