🪛 Installation
You can install the development version of c3t from GitHub with:
if (!require(devtools, quietly = TRUE))
install.packages("devtools")
devtools::install_github("atero18/c3t")📍 Overview
The c3t package in R is a powerful tool for regionalization and clustering tasks with various constraints. It offers a range of functions to facilitate these tasks, and its capabilities include:
1. Hierarchical Regionalization (AHR)
- Solve regionalization problems while respecting constraints such as minimum and maximum size.
- Apply a modified Agglomerative Hierarchical Clustering (AHC) algorithm.
- Choose from various linkage methods to define cluster proximity.
- Apply constraints on the merging process to ensure specific criteria are met.
2. Improving Feasible Solutions
- Enhance feasible solutions while preserving constraints.
- Optimize solutions according to user-defined criteria.
- Explore different linkage methods for clustering.
3. Addressing Unfeasible Solutions
- Convert unfeasible solutions into feasible ones.
- Adjust cluster sizes to meet constraints.
- Maintain contiguity and other specified criteria.
4. Flexible Parameter Customization
- Choose distance measures, such as Euclidean distance.
- Set minimum and maximum size constraints for regions.
- Specify initial partitions for improved results.
- Apply various criteria for evaluation, including Caliński-Harabasz index (CHI) (Caliński and Harabasz 1974).
🚀 Getting Started
Before we begin, ensure you have the c3t package and the required dependencies installed. You can do this by running the following commands:
Creating a Grid
Let’s start by creating a fictitious grid, which will serve as an example for using different functions. We can generate this grid using the gen_grid function. You can specify the number of individuals on the grid, empty cells, and metropolitan areas.
set.seed(123L)
x <- 4L
y <- 5L
nbIndividuals <- 100L
nbEmptyZones <- 3L
nbMetropolises <- 2L
nbVariablesQuant <- 2L
grid <- gen_grid(x, y, nbIndividuals = nbIndividuals,
nbMinEmptyZones = nbEmptyZones,
nbMetropolises = nbMetropolises,
nbQuantitatives = nbVariablesQuant)
data <- grid$context
individuals <- grid$repartition$nbIndividuals
contiguity <- grid$contiguityHierarchical Regionalization (AHR)
To address regionalization problems with constraints like minimum and maximum size, you can use the AHR function. This function applies a modified Hierarchical Agglomerative Clustering (HAC) algorithm while respecting contiguity and size constraints.
resAHR <- AHR(contiguity = contiguity,
d = "euclidean", data = data,
sizes = individuals,
m = 5.0, M = 40.0,
criteria = "CHI",
fusionConstraints = available_fusion_constraints(),
fusionConstraintModes = available_fusion_modes(),
parallel = FALSE)
#> ℹ Starting time: 2023-09-19 22:44:02.865828
#> ℹ 45 AHC to evaluate
#> → 362 non-trivial regionalisations obtained
#> ✔ 62 feasible partitions obtained
#> → 19 redundancies have been removed.
#> → Calculation of the CHI criterion
#> → Execution time: 9.80664587020874The function returns a list of feasible solutions, and you can select the one that best suits your needs.
Improving Feasible Solutions
Once you have a feasible solution, you can further enhance it while preserving constraints using the enhance_feasible function.
resEnhance <-
enhance_feasible(regionalisation = resAHR$results$partition[[1L]],
contiguity = contiguity,
d = "euclidean", data = data,
sizes = individuals,
m = 5.0, M = 40.0,
enhanceCriteria = c("AHC", "CHI"),
linkages = c("single", "complete"),
parallel = FALSE,
verbose = TRUE)
#> → Evaluation of the 3 enhancements
#> → Calculation of 1 evaluation criteria on the initial partition
#> → Calcul of 1 evaluation criteria on the 3 enhanced partitionsThis function allows you to improve your solution according to specified criteria.
Addressing Unfeasible Solutions
In cases where a feasible solution cannot be obtained, the resolve_unfeasible function attempts to provide a feasible solution from an unfeasible one.
unfeasibleReg <- c(1L, 2L, 3L, 2L, 2L, 2L, 2L, 2L, 2L, 4L,
4L, 2L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L)
resolution <- resolve_unfeasible(contiguity = contiguity,
sizes = individuals,
data = data,
d = "euclidean", m = 5.0, M = 40.0,
regionalisation = unfeasibleReg,
verbose = TRUE)
#> → Transfert of elements one-by-one
#> ✔ fully resolved partitionThis function aims to transform an unfeasible solution into a feasible one while respecting constraints.
✅ Conclusion
The c3t package offers a variety of tools for regionalization and clustering with constraints. Explore the documentation and experiment with different functions to suit your specific use case.
For more details on each function and available options, refer to the package documentation :
help(package = "c3t")For any issue or feedback please go to the Issue page of the repository. You can as well discuss about the project in the Discussion page.
Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.