Chapter7.qmd

---
title: "Chapter 7"
subtitle: "Networks"
author: "Aditya Dahiya"
date: "2023-12-09"
format: 
  html: 
    code-fold: true
    code-copy: hover
    code-link: true
execute: 
  echo: true
  warning: false
  error: false
  cache: true
filters:
  - social-share
share:
  permalink: "https://aditya-dahiya.github.io/ggplot2book3e/Chapter7.html"
  description: "Solutions Manual (and Beyond) for ggplot2: Elegant Graphics for Data Analysis (3e)"
  twitter: true
  facebook: true
  linkedin: true
  email: true
  mastodon: true
editor_options: 
  chunk_output_type: console
bibliography: references.bib
---

::: {.callout-note appearance="minimal"}
## Note

There are no exercises in this Chapter to generate solutions for. Instead, we create Network Graphs for Star Wars characters using [Star Wars Social Networks data](https://github.com/evelinag/star-wars-network-data).
:::

```{r}
#| label: setup

# Loading the required libraries
library(tidyverse)
library(ggraph)
library(tidygraph)
library(jsonlite)
```

# 7.1 Network Data

### 7.1.1 Tidygraph: **A tidy network manipulation API**

I demonstrate an example to import a *.json* file using the [Star Wars Social Networks data](https://github.com/evelinag/star-wars-network-data) from [here](https://github.com/evelinag/star-wars-network-data/) [@gabasova2016]. Then, I manipulate the data using `tidygraph` and `tidyverse` functions.

-   Creating a `tbl_graph` object **(`tidygraph`)** [@tidygraph] from a *.json* file using `jsonlite` package. [@jsonlite].

```{r}
#| label: tbl_graph
#| echo: true

# The url for the data
url <- "https://raw.githubusercontent.com/evelinag/star-wars-network-data/master/starwars-full-interactions.json"

# Importing json data
starwars <- jsonlite::read_json(url, simplifyVector = TRUE)

# Getting the nodes (i.e., Star-Wars characters)
sw_nodes <- starwars[[1]] |> 
  as_tibble() |> 
  # An id to match nodes and links
  mutate(id = row_number()) |> 
  relocate(id) |> 
  # Conventional naming
  rename(
    firstname = name,
    name = id
  )

# Getting the links (i.e. connections between characters)
sw_links <- starwars[[2]] |> 
  as_tibble() |> 
  # Network Graphs Tidygraph does not recognize 0 as a node, so adding 1.
  mutate(
    source = source + 1,
    target = target + 1
  ) |> 
  # Conventional Naming
  rename(
    from = source,
    to = target,
    weightage = value
  )

# Creating a tbl_graph object
sw_graph <- tbl_graph(
  nodes = sw_nodes,
  edges = sw_links,
  node_key = "name",
  directed = FALSE
) 

sw_graph
```

### **7.1.3 Algorithms**

The real benefit of tbl_graph data is that we can do many operations on them, while preserving the underlying structure, for example: centrality calculation, ranking, grouping, etc.

Finding groups within Star Wars Characters using `group_leading_eigen()` which groups nodes based on the leading eigenvector of the modularity matrix using `igraph::cluster_leading_eigen()` . The variable `col_var` shows the group, which can be used to create groups.

```{r}
#| code-fold: false

sw_graph <- sw_graph |> 
  # Creating groups with group_leading_eigen()
  activate(nodes) |> 
  mutate(col_var = tidygraph::group_leading_eigen())
```

The `group_leading_eigen()` creates groups, and as we can see below in @fig-eigen , the groups are of related characters who appear often together in the movie.

```{r}
#| label: fig-eigen
#| fig-cap: "Groups of Star Wars characters as formed by group_leading_eigen()"

sw_graph |> 
  activate(nodes) |> 
  as_tibble() |> 
  select(col_var, firstname, value) |> 
  mutate(
    Group = paste0("Gp. ", col_var),
    firstname = snakecase::to_title_case(firstname)
  ) |> 
  group_by(Group) |>
  arrange(desc(value)) |> 
  summarise(Characters = paste(firstname, collapse = ", ")) |> 
  gt::gt() |> 
  gtExtras::gt_theme_espn()
```

# 7.2 Visualizing Networks

We can use the `tbl_graph` object with `ggraph` to create visualizations easily.

-   In @fig-v1, we are visualizing Networks amongst most important characters (imporantance determined by \>40 appearances in scenes). As we can see, out `tidygraph` function `group_leading_eigen()` has successfully group the characters into groups, based on their interactions.

```{r}
#| label: fig-v1
#| fig-cap-location: top
#| fig-cap: "Linkages (defined by simultaneous appearance in a scene in the movie) amongst most important characters of Star Wars movies"

sw_graph |> 
  
  activate(nodes) |> 
  filter(value > 40) |> 
  
  ggraph(layout = "stress") +
  geom_edge_link(
    mapping = aes(width = weightage),
    alpha = 0.5,
    color = "grey"
  ) +
  geom_node_point(
    mapping = aes(size = value, 
                  colour = as.factor(col_var))
  ) +
  geom_node_text(
    mapping = aes(label = firstname),
    check_overlap = TRUE,
    repel = TRUE
  ) +
  scale_size_area(max_size = 10) +
  scale_colour_brewer(palette = "Dark2") +
  theme_void() + 
  theme(legend.position = "none")
```

-   Below, we trying out different layouts in @fig-layouts using the argument `layout = ""` to the function `ggraph()`. The `ggraph` provides well over 20 different layouts[^1] to choose from, including (Credits: [`R` Documentation](https://www.rdocumentation.org/packages/ggraph/versions/2.1.0/topics/layout_tbl_graph_igraph) for `layout_tbl_graph_igraph`): ---

[^1]: The layouts have been picked up (copied) from [RDocumentation](https://www.rdocumentation.org/packages/ggraph/versions/2.1.0/topics/layout_tbl_graph_igraph) webpage: `ggraph` (version 2.1.0) `layout_tbl_graph_igraph`: using `igraph` layout algorithms for `layout_tbl_graph`

<!-- -->

-   **Hierarchical layouts**

    -   **`tree`**: Uses the Reingold-Tilford algorithm to place the nodes below their parent with the parent centered above its children.

    -   **`sugiyama`**: Designed for directed acyclic graphs (that is, hierarchies where multiple parents are allowed) it minimizes the number of crossing edges.

-   **Standard layouts**

    -   **`bipartite`**: Minimize edge-crossings in a simple two-row (or column) layout for bipartite graphs.

    -   **`star`**: Place one node in the center and the rest equidistantly around it.

    -   **`circle`**: Place nodes in a circle in the order of their index. Consider using layout_tbl_graph_linear() with circular=TRUE for more control.

    -   **`nicely`**: Tries to pick an appropriate layout. See igraph::nicely() for a description of the simple decision tree it uses

    -   **`dh`**: Uses Davidson and Harels simulated annealing algorithm to place nodes.

    -   **`gem`**: Place nodes on the plane using the GEM force-directed layout algorithm.

    -   **`graphopt`**: Uses the Graphopt algorithm based on alternating attraction and repulsion to place nodes.

    -   **`grid`**: Place nodes on a rectangular grid.

    -   **`mds`**: Perform a multidimensional scaling of nodes using either the shortest path or a user supplied distance.

    -   **`sphere`**: Place nodes uniformly on a sphere - less relevant for 2D visualizations of networks.

    -   **`randomly`**: Places nodes uniformly random.

    -   **`fr`**: Places nodes according to the force-directed algorithm of Fruchterman and Reingold.

    -   **`kk`**: Uses the spring-based algorithm by Kamada and Kawai to place nodes.

    -   **`drl`**: Uses the force directed algorithm from the DrL toolbox to place nodes.

    -   **`lgl`**: Uses the algorithm from Large Graph Layout to place nodes.

```{r}
#| label: fig-layouts
#| layout-ncol: 2
#| fig-asp: 1
#| fig-cap: "Different Layouts for the ggraph()"
#| fig-subcap: 
#|   - "drl"
#|   - "circle"
#|   - "nicely"
#|   - "dh"
#|   - "gem"
#|   - "graphopt"
#|   - "grid"
#|   - "kk"

sw_graph_fun <- function(my_layout){
  sw_graph |>
  activate(nodes) |>
  filter(value > 40) |> 
  ggraph(layout = my_layout) +
  geom_edge_link(
    mapping = aes(width = weightage),
    alpha = 0.5,
    color = "grey"
  ) +
  geom_node_point(mapping = aes(size = value, 
                                col = as.factor(col_var))) +
  geom_node_text(
    mapping = aes(label = firstname),
    check_overlap = TRUE, 
    repel = TRUE,
    fontface = "bold"
  ) +
  labs(title = paste0("layout = \"", my_layout, "\"")) +
  scale_size_area(max_size = 10) +
  theme_void() +
  theme(
    legend.position = "none",
    plot.title = element_text(
      face = "bold",
      family = "mono",
      size = 20,
      hjust = 0
    )
  )
}

sw_graph_fun("drl")

sw_graph_fun("circle")

sw_graph_fun("nicely")

sw_graph_fun("dh")

sw_graph_fun("gem")

sw_graph_fun("graphopt")

sw_graph_fun("grid")

sw_graph_fun("kk")
```

-   We can further customizing Edge-widths, curvature and colours to represent strength of links between the nodes using `ggraph`, as depicted in @fig-edges.

```{r}
#| label: fig-edges
#| fig-cap: "Customizing the links between the nodes with geom_edge_link2()"

library(showtext)
library(fontawesome)

font_add_google("Poller One",
                family = "font_sw")

font_add_google("PT Sans Narrow", "font_title")
font_add_google("Nova Square", "font_body")
text_col <- "black"
# Caption stuff
github <- "&#xf09b"
github_username <- "aditya-dahiya"
xtwitter <- "&#xe61b"
xtwitter_username <- "adityadahiyaias"
linkedin <- "&#xf08c"
linkedin_username <- "dr-aditya-dahiya-ias"
social_caption <- glue::glue("<span style='font-family:\"Font Awesome 6 Brands\";'>{github};</span> <span style='color: {text_col}'>{github_username}  </span> <span style='font-family:\"Font Awesome 6 Brands\";'>{xtwitter};</span> <span style='color: {text_col}'>{xtwitter_username}</span> <span style='font-family:\"Font Awesome 6 Brands\";'>{linkedin};</span> <span style='color: {text_col}'>{linkedin_username}</span>")


showtext_auto()
set.seed(4)
sw_graph |>
  activate(nodes) |>
  mutate(col_var = as.character(col_var)) |>
  filter(value > 40) |>
  
  # Start plotting network graph
  ggraph(layout = "nicely") +
  
  # Edges
  geom_edge_bend2(
    aes(
      colour = node.col_var,
      width = weightage
    ),
    lineend = "round"
  ) +
  
  # Nodes
  geom_node_point(
    aes(
      size = value,
      fill = col_var
    ),
    pch = 21,
    colour = "white",
    stroke = 2
  ) +
  
  # Labels
  geom_node_text(
    aes(label = firstname),
    check_overlap = TRUE,
    repel = TRUE,
    fontface = "bold"
  ) +
  labs(title = "Network showing appearances of main characters in Star Wars films",
       subtitle = "Size of dot indicates screen-time; width of link indicates common appearances in scenes",
       caption = social_caption) +
  scale_size_area(max_size = 15) +
  theme_void() +
  theme(
    legend.position = "none",
    plot.title = element_text(
      family = "font_title",
      face = "bold",
      size = 21,
      hjust = 0
    ),
    plot.subtitle = element_text(
      family = "font_body",
      hjust = 0
    ),
    plot.caption = element_text(
      hjust = 1,
      family = "font_title"
    )
  )
```

-   **Faceting:** Lastly, we can also facet on the nodes or on the edges, and produce multiple network graphs, as shown in @fig-facet for the entire Star Wars cast (not just the most important characters): ---

```{r}
#| label: fig-facet
#| fig-width: 10
#| fig-asp: 1.5
#| fig-cap: "Different Network of Characters in Star Wars: A Faceted ggraph"

labels_facet = c(
  "Jedi & Sith Lords",
  "Prequel Trilogy", 
  "Original Trilogy",
  "Sequel Trilogy",
  "Villains",
  "An outlier!"
)
labels_facet <- str_to_upper(labels_facet)
names(labels_facet) = 1:6

sw_graph |>
  activate(nodes) |> 
  mutate(firstname = snakecase::to_title_case(firstname)) |> 
  ggraph() +
  geom_edge_link(
    mapping = aes(width = weightage),
    alpha = 0.5,
    color = "grey"
  ) +
  geom_node_point(
    aes(size = value, 
        col = as.factor(col_var))) +
  geom_node_text(
    aes(label = firstname,
        size = value),
    repel = TRUE
  ) +
  scale_size_area(max_size = 10) +
  scale_size_continuous(range = c(6, 15)) +
  scale_color_brewer(palette = "Set2") +
  facet_nodes(~ col_var, 
              scales = "free",
              labeller = as_labeller(labels_facet),
              ncol = 2) +
  theme_void() +
  theme(
    legend.position = "none", 
    strip.text = element_text(
      family = "font_sw",
      hjust = 0.5,
      size = 36
    ),
    panel.border = element_rect(colour = "darkgrey",
                                fill = NA),
    strip.background = element_rect(colour = "darkgrey",
                                    fill = NA)
  )
```

```{r}
#| label: Code-for-social-media-post
#| eval: false
#| echo: false

library(showtext)
library(fontawesome)
sysfonts::font_add(family = "Font Awesome 6 Brands",
                   regular = "C:/Users/dradi/Documents/Post-Harvard Academics/projects_presentations/docs/Font Awesome 6 Brands-Regular-400.otf")
  
font_add_google("Poller One",
                family = "font_sw")

font_add_google("PT Sans Narrow", "font_title")
font_add_google("Nova Square", "font_body")
text_col <- "black"
# Caption stuff
github <- "&#xf09b"
github_username <- "aditya-dahiya"
xtwitter <- "&#xe61b"
xtwitter_username <- "adityadahiyaias"
linkedin <- "&#xf08c"
linkedin_username <- "dr-aditya-dahiya-ias"
social_caption <- glue::glue("<span style='font-family:\"Font Awesome 6 Brands\";'>{github};</span> <span style='color: {text_col}'>{github_username}  </span> <span style='font-family:\"Font Awesome 6 Brands\";'>{xtwitter};</span> <span style='color: {text_col}'>{xtwitter_username}</span> <span style='font-family:\"Font Awesome 6 Brands\";'>{linkedin};</span> <span style='color: {text_col}'>{linkedin_username}</span>")

library(magick)
img_sw <- image_read("https://www.freepnglogos.com/uploads/star-wars-logo-31.png")

showtext_auto()
set.seed(4)
g <- sw_graph |>
  activate(nodes) |>
  mutate(col_var = as.character(col_var)) |>
  filter(value > 40) |>
  
  # Start plotting network graph
  ggraph(layout = "nicely") +
  
  # Edges
  geom_edge_bend2(
    aes(
      colour = node.col_var,
      width = weightage
    ),
    lineend = "round"
  ) +
  
  # Nodes
  geom_node_point(
    aes(
      size = value,
      fill = col_var
    ),
    pch = 21,
    colour = "white",
    stroke = 2
  ) +
  
  # Labels
  geom_node_text(
    aes(label = firstname),
    check_overlap = TRUE,
    repel = TRUE,
    fontface = "bold"
  ) +
  theme_classic() +
  labs(title = "Network showing appearances of main characters in Star Wars films",
       subtitle = "Size of dot indicates screen-time; width of link indicates common appearances in scenes",
       caption = social_caption) +
  scale_size_area(max_size = 15) +
  theme_void() +
  theme(
    legend.position = "none",
    plot.title = element_text(
      family = "font_title",
      face = "bold",
      size = 21,
      hjust = 0
    ),
    plot.subtitle = element_text(
      family = "font_body",
      hjust = 0
    ),
    plot.caption = ggtext::element_textbox(
      hjust = 0.1,
      family = "font_title"
    )
  )

g +
  annotation_raster(img_sw,
                    xmin = -1.2,
                    xmax = 0,
                    ymin = -3,
                    ymax = -1.5)

```