Spatially Constrained Clustering of Nigerian States: Perspective from Social, Economic and Demographic Attributes

摘要

Creationand differentiation of regions are some of the basic tasks in geographicanalysis. Regionalisation attempts to create a generalised representation ofthe processes which is taking place at the level of the amalgamated geographicunits. To this end, this study examined the combined use of demographic,economic and poverty characteristics of States across Nigeria to create regionsrelevant for economic and development planning. The study utilised dependencyratios derived from gridded age structure data, Gross domestic product (GDP),poverty index. K-Means and Max-p algorithm were used for identification ofregions. Correlation analysis showed thatYouth dependency and total dependency have a strong statistically significantpositive relationship (r=0.998, p<0.01) indicating that dependency in thecountry is driven by youth. The best K-Mean clustering implementation withoutconsidering contiguity identified 12 regions with a ratio of between and totalsum of squares (RBTSS) of 0.789. The Max-p algorithm was tested with populationconstrain, the best result identified 9 regions with RBTSS of 0.611 constrainedby a minimum population of 8% and implemented with the greedy local searchalgorithm, this was the same for the simulated annealing approach (SA). Withhigh dissimilarity still common across a handful of the regions identified, afurther test was carried out using a minimum bound of 3 States and the SA localsearch approach. The best result identified 11 contiguous regions with only oneregion having a relatively high within region dissimilarity and a RBTSS of0.626. The results confirmed that there are more than 6 regions as currentlydefined for the country. The analyses showcased an example of knowledgediscovery from a spatial dataset which could support regional developmentplanning. From the results, there is a clear need for re-examination of currentregions and designing of better-defined regions to ensure that development isguided by evidence.