In pachamaltese/binet: Functions to build bipartite networks under the Tidyverse philosophy

knitr::opts_chunk$set(eval = TRUE, message = FALSE, warning = FALSE)

Galactic Federation

This example is inspired after the visualizations from @atlas2014 but heavily inspired after Rick and Morty to keep it simple and lightweight.

Package data

I'll use the Galactic Federation dataset to illustrate the essential usage of the functions within this package.

library(binet)

galactic_federation

Balassa Index

You can obtain Balassa Index with balassa_index().

bi <- balassa_index(
  data = galactic_federation,
  source = "planet",
  target = "product",
  value = "export_value"
)

bi

Another possibility is to obtain Balassa Index without discretization.

bi_dec <- balassa_index(
  data =galactic_federation,
  source = "planet",
  target = "product",
  value = "export_value",
  discrete = F
)

bi_dec

Complexity Measures

You can compute complexity indexes (e.g. such as the Economic Complexity Index and Product Complexity Index) by using complexity_measures(). The calculations methods are fitness (default), reflections, eigenvalues. See [@measuringcomplexity2015] for the methodological details.

The eigenvalues also calls the reflections methods in order to correct the index sign in some special cases when the correlation between the output from both methods is negative.

Complexity-Fitness

com_fit <- complexity_measures(balassa_index = bi)

com_fit$complexity_index_source
com_fit$complexity_index_target

Reflections

com_ref <- complexity_measures(
  balassa_index = bi,
  method = "reflections"
)

com_ref$complexity_index_source
com_ref$complexity_index_target

Eigenvalues

com_eig <- complexity_measures(
  balassa_index = bi,
  method = "eigenvalues"
)

com_eig$complexity_index_source
com_eig$complexity_index_target

Proximity

Proximity matrices are used to create projections e.g. (country-country and product-product networks) for bipartite networks. Using proximity() is straightforward.

pro <- proximity(
  balassa_index = bi,
  balassa_sum_source = com_fit$balassa_sum_source,
  balassa_sum_target = com_fit$balassa_sum_target
)

pro$proximity_source
pro$proximity_target

Networks

The projections() function is designed to use igraph for the internal computations and also to pass proximity-based networks to igraph, ggraph or export to Cytoscape by saving the output as csv/tsv.

net <- projections(
  proximity_source = pro$proximity_source,
  proximity_target = pro$proximity_target,
  avg_links = 4,
  tolerance = 0.05
)

net$network_source
net$network_target

Just two basic examples with ggraph.

set.seed(200100)
library(Matrix)
library(igraph)
library(ggraph)

aggregated_planets <- rowSums(galactic_federation)
V(net$network_source)$size <- aggregated_planets[match(V(net$network_source)$name, names(aggregated_planets))]

net$network_source %>% 
  ggraph(layout = "kk") +
  geom_edge_link(aes(edge_width = weight), edge_colour = "#a8a8a8") +
  geom_node_point(aes(size = size), color = "darkslategray4") +
  geom_node_text(aes(label = name), vjust = 2.2) +
  ggtitle("Proximity Based Network Projection for Planets") +
  theme_void()

The network associated to

set.seed(200100)

aggregated_products <- colSums(galactic_federation)
V(net$network_target)$size <- aggregated_products[match(V(net$network_target)$name, names(aggregated_products))]

net$network_target %>% 
  ggraph(layout = "kk") +
  geom_edge_link(aes(edge_width = weight), edge_colour = "#a8a8a8") +
  geom_node_point(aes(size = size), color = "darkslategray4") +
  geom_node_text(aes(label = name), vjust = 2.2) +
  ggtitle("Proximity Based Network Projection for Products") +
  theme_void()