Hierarchical Clustering with Contiguity Constraint in <i>R</i>

This article presents a new implementation of hierarchical clustering for the R language that allows one to apply spatial or temporal contiguity constraints during process. The need constraint arises, instance, when wants partition map into different domains similar physical conditions, identify discontinuities in time series, group regional administrative units with respect their performance, and so on. To increase computation efficiency, we programmed core functions plain C. result is function, constr.hclust, which distributed package adespatial. program implements general agglomerative algorithm described by Lance Williams (1966; 1967), particularity allowing only clusters are contiguous geographic space along fuse at any given step. Contiguity can be defined time. Information about provided connection network among sites, edges describing links between connected sites. Clustering also known as chronological clustering. on implicitly rank positions observations series. was mirrored found function hclust standard stats (R Core Team 2022). We transcribed from Fortran C added functionality running function. efficient. It limited mainly input/output access massive amounts memory potentially needed store copies dissimilarity matrix update its elements analyzing large problems. computer code plotting results numbers clusters.