To build this prototype module, we adopted a structured visual analytics approach, starting from data ingestion to visual storytelling. The knowledge graph provided in Mini-Challenge 3 was first cleaned, transformed, and modeled using tidygraph. Communication patterns, relationships, and anomalies were then extracted through modular analysis and visualized using both static (ggplot2, ggraph) and interactive (plotly, visNetwork) tools.
The overall methodology consists of the following stages:
Data Preparation We loaded and flattened the JSON-based knowledge graph, extracted nodes and edges, cleaned missing values, and validated the network structure for compatibility with tidygraph. Each entity was classified by type and sub-type to enable filtering and targeted analysis.
Modular Task Execution Each of the four investigative tasks was assigned to team member, allowing parallel and comparison development:
Prototype Development All modules were translated into reactive components suitable for a Shiny application. Input widgets such as selectInput(), dateRangeInput(), and checkboxGroupInput() allow user-driven filtering, while outputs include plotlyOutput() and visNetworkOutput() for dynamic and exploratory analysis.
Design and Storyboarding A UI storyboard was drafted to guide the layout of the final Shiny app, ensuring usability and clarity. Each module was built to function independently but integrate smoothly into the full application.
We will use the code chunk below uses p_load() of pacman package to check if packages are installed in the computer. If they are, then they will be launched into R. Though it is not added here in pacman due to conflicts, networkD3, highcharter, leaflet, conflicted will also be added on Shiny.
pacman::p_load(jsonlite, tidyverse, ggtext,
knitr, lubridate, hms, scales,
tidytext, tm, SnowballC,
patchwork, ggraph,
tidygraph, igraph, ggiraph,
SmartEDA, plotly, wordcloud,
ggh4x, visNetwork, RColorBrewer,
circlize, ggalluvial, reactable) In the code chunk below, fromJSON() of jsonlite package was used to import mc3.json file into R and save the output object.
mc3_data <- fromJSON("data/mc3_graph.json")
mc3_schema <- fromJSON("data/MC3_schema.json")It contains graph data, where nodes can be accessed via nodes and edges via links. This dataset had many columns but we filtered the relevant data during wrangling.
We will also set some values for consistency throughout all graphs.
node_legend_colors_plot <- c(
"Person" = "#88CCEE",
"Vessel" = "#D55E00",
"Organization" = "#117733",
"Location" = "#AA4499",
"Group"= "#CC79A7",
"Event" = "#DDCC77",
"Relationship" = "#AF8DC3",
"Nadia Conti" = "red"
)
node_legend_shapes_plot <- c(
"Person" = "dot",
"Vessel" = "triangle",
"Organization" = "square",
"Location" = "diamond",
"Group" = "circle plus",
"Event" = "star",
"Relationship" = "square x",
"Nadia Conti" = "star"
)
STYLES <- list(
node_label_dark = "black",
font_family = "Roboto Condensed"
)In the code chunk below glimpse() is used to reveal the structure of mc3_data knowledge graph.
glimpse(mc3_data)List of 5
$ directed : logi TRUE
$ multigraph: logi FALSE
$ graph :List of 4
..$ mode : chr "static"
..$ edge_default: Named list()
..$ node_default: Named list()
..$ name : chr "VAST_MC3_Knowledge_Graph"
$ nodes :'data.frame': 1159 obs. of 31 variables:
..$ type : chr [1:1159] "Entity" "Entity" "Entity" "Entity" ...
..$ label : chr [1:1159] "Sam" "Kelly" "Nadia Conti" "Elise" ...
..$ name : chr [1:1159] "Sam" "Kelly" "Nadia Conti" "Elise" ...
..$ sub_type : chr [1:1159] "Person" "Person" "Person" "Person" ...
..$ id : chr [1:1159] "Sam" "Kelly" "Nadia Conti" "Elise" ...
..$ timestamp : chr [1:1159] NA NA NA NA ...
..$ monitoring_type : chr [1:1159] NA NA NA NA ...
..$ findings : chr [1:1159] NA NA NA NA ...
..$ content : chr [1:1159] NA NA NA NA ...
..$ assessment_type : chr [1:1159] NA NA NA NA ...
..$ results : chr [1:1159] NA NA NA NA ...
..$ movement_type : chr [1:1159] NA NA NA NA ...
..$ destination : chr [1:1159] NA NA NA NA ...
..$ enforcement_type : chr [1:1159] NA NA NA NA ...
..$ outcome : chr [1:1159] NA NA NA NA ...
..$ activity_type : chr [1:1159] NA NA NA NA ...
..$ participants : int [1:1159] NA NA NA NA NA NA NA NA NA NA ...
..$ thing_collected :'data.frame': 1159 obs. of 2 variables:
.. ..$ type: chr [1:1159] NA NA NA NA ...
.. ..$ name: chr [1:1159] NA NA NA NA ...
..$ reference : chr [1:1159] NA NA NA NA ...
..$ date : chr [1:1159] NA NA NA NA ...
..$ time : chr [1:1159] NA NA NA NA ...
..$ friendship_type : chr [1:1159] NA NA NA NA ...
..$ permission_type : chr [1:1159] NA NA NA NA ...
..$ start_date : chr [1:1159] NA NA NA NA ...
..$ end_date : chr [1:1159] NA NA NA NA ...
..$ report_type : chr [1:1159] NA NA NA NA ...
..$ submission_date : chr [1:1159] NA NA NA NA ...
..$ jurisdiction_type: chr [1:1159] NA NA NA NA ...
..$ authority_level : chr [1:1159] NA NA NA NA ...
..$ coordination_type: chr [1:1159] NA NA NA NA ...
..$ operational_role : chr [1:1159] NA NA NA NA ...
$ edges :'data.frame': 3226 obs. of 5 variables:
..$ id : chr [1:3226] "2" "3" "5" "3013" ...
..$ is_inferred: logi [1:3226] TRUE FALSE TRUE TRUE TRUE TRUE ...
..$ source : chr [1:3226] "Sam" "Sam" "Sam" "Sam" ...
..$ target : chr [1:3226] "Relationship_Suspicious_217" "Event_Communication_370" "Event_Assessment_600" "Relationship_Colleagues_430" ...
..$ type : chr [1:3226] NA "sent" NA NA ...Next, as_tibble() of tibble package package is used to extract the nodes and links tibble data frames from mc3 tibble dataframe into two separate tibble dataframes called mc3_nodes_raw and mc3_edges_raw respectively.
mc3_nodes_raw <- as_tibble(mc3_data$nodes)
mc3_edges_raw <- as_tibble(mc3_data$edges)We also looked into the nodes and edges structure.
ExpData(data=mc3_nodes_raw,type=2) Index Variable_Name Variable_Type Sample_n Missing_Count Per_of_Missing
1 1 type character 1159 0 0.000
2 2 label character 1159 0 0.000
3 3 name character 72 1087 0.938
4 4 sub_type character 1159 0 0.000
5 5 id character 1159 0 0.000
6 6 timestamp character 770 389 0.336
7 7 monitoring_type character 70 1089 0.940
8 8 findings character 70 1089 0.940
9 9 content character 584 575 0.496
10 10 assessment_type character 33 1126 0.972
11 11 results character 32 1127 0.972
12 12 movement_type character 33 1126 0.972
13 13 destination character 41 1118 0.965
14 14 enforcement_type character 21 1138 0.982
15 15 outcome character 19 1140 0.984
16 16 activity_type character 4 1155 0.997
17 17 participants integer 1 1158 0.999
18 18 thing_collected data.frame -2314 2316 1158.000
19 19 reference character 1 1158 0.999
20 20 date character 5 1154 0.996
21 21 time character 3 1156 0.997
22 22 friendship_type character 2 1157 0.998
23 23 permission_type character 55 1104 0.953
24 24 start_date character 90 1069 0.922
25 25 end_date character 6 1153 0.995
26 26 report_type character 19 1140 0.984
27 27 submission_date character 22 1137 0.981
28 28 jurisdiction_type character 13 1146 0.989
29 29 authority_level character 7 1152 0.994
30 30 coordination_type character 73 1086 0.937
31 31 operational_role character 38 1121 0.967
No_of_distinct_values
1 3
2 92
3 73
4 25
5 1159
6 612
7 6
8 70
9 581
10 4
11 33
12 4
13 26
14 3
15 20
16 3
17 1
18 2
19 2
20 6
21 4
22 3
23 4
24 69
25 6
26 4
27 23
28 13
29 3
30 53
31 35ExpData(data=mc3_edges_raw,type=2) Index Variable_Name Variable_Type Sample_n Missing_Count Per_of_Missing
1 1 id character 3169 57 0.018
2 2 is_inferred logical 3226 0 0.000
3 3 source character 3226 0 0.000
4 4 target character 3226 0 0.000
5 5 type character 2204 1022 0.317
No_of_distinct_values
1 3170
2 2
3 1052
4 1156
5 4In the code chunk below, ExpCatViz() of SmartEDA package is used to reveal the frequency distribution of all categorical fields in mc3_nodes tibble dataframe.
library(shiny)
library(ggplot2)
# Run ExpCatViz once at the top to avoid recomputing
ExpCatViz(data=mc3_nodes_raw,
col="navyblue")[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
[[9]]
[[10]]
[[11]]
[[12]]
[[13]]
[[14]]
Code chunk below uses ggplot2 functions to reveal the frequency distribution of sub_type field of mc3_nodes_raw.
# Step 1: Count and reorder
mc3_nodes_ordered <- mc3_nodes_raw %>%
count(sub_type) %>%
arrange((n)) %>%
mutate(sub_type = factor(sub_type, levels = sub_type))
# Step 2: Plot with navy bars, sorted, and horizontal
ggplot(mc3_nodes_ordered, aes(x = sub_type, y = n)) +
geom_col(fill = "navy") +
coord_flip() +
labs(x = "Sub_type", y = "Count",
title = "Distribution of Subtypes") +
theme_minimal()
In the code chunk below, the Entity subtypes are filtered.
# Step 1: Filter for type == "Entity", count sub_type, sort
relationship_subtypes <- mc3_nodes_raw %>%
filter(type == "Entity") %>%
count(sub_type) %>%
arrange(n) %>%
mutate(sub_type = factor(sub_type, levels = sub_type))
# Step 2: Plot
ggplot(relationship_subtypes, aes(x = sub_type, y = n)) +
geom_col(fill = "navy") +
coord_flip() +
labs(
x = "Entity Subtype",
y = "Count",
title = "Distribution of Entity Subtypes"
) +
theme_minimal()
In the code chunk below, the Event subtypes are filtered.
# Step 1: Filter for type == "Event", count sub_type, sort
relationship_subtypes <- mc3_nodes_raw %>%
filter(type == "Event") %>%
count(sub_type) %>%
arrange(n) %>%
mutate(sub_type = factor(sub_type, levels = sub_type))
# Step 2: Plot
ggplot(relationship_subtypes, aes(x = sub_type, y = n)) +
geom_col(fill = "navy") +
coord_flip() +
labs(
x = "Event Subtype",
y = "Count",
title = "Distribution of Event Subtypes"
) +
theme_minimal()
In the code chunk below, the relationship subtypes are filtered.
# Step 1: Filter for type == "Relationship", count sub_type, sort
relationship_subtypes <- mc3_nodes_raw %>%
filter(type == "Relationship") %>%
count(sub_type) %>%
arrange(n) %>%
mutate(sub_type = factor(sub_type, levels = sub_type))
# Step 2: Plot
ggplot(relationship_subtypes, aes(x = sub_type, y = n)) +
geom_col(fill = "navy") +
coord_flip() +
labs(
x = "Relationship Subtype",
y = "Count",
title = "Distribution of Relationship Subtypes"
) +
theme_minimal()
We will use the EDA findings to determine data to focus on or eliminate. From the bar charts and the original data on mc3_nodes_raw, it was observed that:
Nodes were one of three types (Entity, Event, Relationship), where each of these types have their sub_types. Majority were of event type, followed by relationship, and entity.
Observations of EDA from Event types:
Findings field were filled when there were monitoring_type.
Content refers to radio communication content.
Results field were filled when there were assessment_type performed.
When there is an enforcement_type of enforcement operations or warnings, there might be an outcome at times.
When there is a movement_type, there might be a place of destination at times.
Observations of EDA from Relationship types:
When the subtype was coordinate, there were data in the field named coordination_types.
When the subtype was operate, there were data in the field named operational_roles.
When there is a jurisdiction_type, there might be an authority_level.
There are only restricted or special access data within permission_types.
When there is a report_type of data transmission or environmental report, there might be a submission_date.
Observations of EDA from Entity types:
Elimination and directed focus:
Relative to the entire dataset, there were little assessment_type (3%), movement_type (2%), enforcement_type (2%), permission_type (4%), report_type (2%), authority_level (1%). We will direct our focus on other areas instead of these.
There were no to little useful data in the fields named: activity_type, references, dates, time, and friendship_type. These were not utilised.
We directed our focus on Event_Communication, Event_Monitoring, and Event_VesselMovement.
The code chunk below used ExpCATViz() of SmartEDA package to reveal the frequency distribution of all categorical fields in mc3_edges_raw tibble dataframe.
ExpCatViz(data=mc3_edges_raw,
col="navyblue")[[1]]
Entities are connected by edges to other Entities via an Event or Relationship node. The one exception to this is the Communication Event subtype, which is additionally linked to either an Event or Relationship node. The type field denotes the connector or edge type for the Entities, Event, and Relationship nodes. The edges are one of these: received, evidence_for, sent, NA.
# Step 1: Filter for type == "sent"
filtered_edges <- mc3_edges_raw %>%
filter(type == "sent") %>%
count(source) %>%
arrange(desc(n)) %>%
mutate(source = factor(source, levels = rev(unique(source)))) # descending
# Step 2: Plot
ggplot(filtered_edges, aes(x = source, y = n)) +
geom_col(fill = "navy") +
coord_flip() +
labs(
title = "Distribution of 'sent' Edges type by Source",
x = "Source",
y = "Count"
) +
theme_minimal()
What we understood from the information provided by Vast Challenge on Directional Edges:
For relationship as colleagues node or friends node, the node will have arrows/ edges pointing towards the relationship node.
For other relationships and events, the direction would be following the source and target.
convert values in id field into character data type,
exclude records with id value are na,
exclude records with similar id values,
exclude thing_collected , time , date, friendship_type field, and
save the cleaned tibble dataframe into a new tibble datatable called mc3_nodes_cleaned.
mc3_nodes_cleaned <- mc3_nodes_raw %>%
mutate(id = as.character(id)) %>%
filter(!is.na(id)) %>%
distinct(id, .keep_all = TRUE) %>%
select(-thing_collected, -time, -date, -friendship_type)# A tibble: 27 × 2
column unique_count
<chr> <int>
1 id 1159
2 timestamp 612
3 content 581
4 label 92
5 name 73
6 findings 70
7 start_date 69
8 coordination_type 53
9 operational_role 35
10 results 33
# ℹ 17 more rowsrenamed source and target fields to from_id and to_id respectively,
converted values in from_id and to_id fields to character data type,
excluded values in from_id and to_id which not found in the id field of mc3_nodes_cleaned,
excluded records whereby from_id and/or to_id values are missing, and
saved the cleaned tibble dataframe and called it mc3_edges_cleaned.
mc3_edges_cleaned <- mc3_edges_raw %>%
rename(from_id = source,
to_id = target) %>%
mutate(across(c(from_id, to_id), as.character)) %>%
# Parse to_id to get supertype and sub_type for target nodes (e.g., Event_Communication)
separate(to_id, into = c("to_id_supertype", "to_id_sub_type", "to_id_num"),
sep = "_", remove = FALSE, fill = "right", extra = "merge") %>%
# Filter to ensure from_id and to_id exist in mc3_nodes_cleaned (prevent orphaned edges)
filter(from_id %in% mc3_nodes_cleaned$id,
to_id %in% mc3_nodes_cleaned$id) %>%
filter(!is.na(from_id), !is.na(to_id))
print("Columns in mc3_edges_cleaned after initial cleaning:")[1] "Columns in mc3_edges_cleaned after initial cleaning:"print(colnames(mc3_edges_cleaned))[1] "id" "is_inferred" "from_id" "to_id"
[5] "to_id_supertype" "to_id_sub_type" "to_id_num" "type" print("Head of mc3_edges_cleaned after initial cleaning:")[1] "Head of mc3_edges_cleaned after initial cleaning:"print(head(mc3_edges_cleaned))# A tibble: 6 × 8
id is_inferred from_id to_id to_id_supertype to_id_sub_type to_id_num type
<chr> <lgl> <chr> <chr> <chr> <chr> <chr> <chr>
1 2 TRUE Sam Rela… Relationship Suspicious 217 <NA>
2 3 FALSE Sam Even… Event Communication 370 sent
3 5 TRUE Sam Even… Event Assessment 600 <NA>
4 3013 TRUE Sam Rela… Relationship Colleagues 430 <NA>
5 <NA> TRUE Sam Rela… Relationship Friends 272 <NA>
6 <NA> TRUE Sam Rela… Relationship Colleagues 215 <NA> # Find the number of unique types in each column
unique_counts <- mc3_edges_cleaned %>%
summarise_all(n_distinct) %>%
pivot_longer(cols = everything(), names_to = "column", values_to = "unique_count")
# Print the unique counts for each column
print(unique_counts)# A tibble: 8 × 2
column unique_count
<chr> <int>
1 id 3170
2 is_inferred 2
3 from_id 1052
4 to_id 1156
5 to_id_supertype 71
6 to_id_sub_type 21
7 to_id_num 860
8 type 4Next, code chunk below will be used to create mapping of character id in mc3_nodes_cleaned to row index
node_index_lookup <- mc3_nodes_cleaned %>%
mutate(.row_id = row_number()) %>%
select(id, .row_id)Next, the code chunk below was used to join and convert from_id and to_id to integer indices. At the same time we also dropped rows with unmatched nodes.
mc3_edges_indexed <- mc3_edges_cleaned %>%
left_join(node_index_lookup, by = c("from_id" = "id")) %>%
rename(from = .row_id) %>%
left_join(node_index_lookup, by = c("to_id" = "id")) %>%
rename(to = .row_id) %>%
# Filter out edges where either source or target node was not found
filter(!is.na(from) & !is.na(to)) %>%
# Select all columns to carry forward to mc3_edges_final
select(from, to, id, is_inferred, type, # Original edge attributes
from_id, to_id, to_id_supertype, to_id_sub_type, to_id_num # Original IDs and parsed target type
)Next the code chunk below was used to subset nodes to only those referenced by edges.
used_node_indices <- sort(unique(c(mc3_edges_indexed$from, mc3_edges_indexed$to)))
mc3_nodes_final <- mc3_nodes_cleaned %>%
slice(used_node_indices) %>%
mutate(new_index = row_number())We then used the code chunk below to rebuild lookup from old index to new index.
old_to_new_index <- tibble(
old_index = used_node_indices,
new_index = seq_along(used_node_indices)
)Lastly, the code chunk below was used to update edge indices to match new node table.
mc3_edges_final <- mc3_edges_indexed %>%
left_join(old_to_new_index, by = c("from" = "old_index")) %>%
rename(from_new = new_index) %>%
left_join(old_to_new_index, by = c("to" = "old_index")) %>%
rename(to_new = new_index) %>%
# Explicitly select all columns that are needed downstream
select(from = from_new, to = to_new,
id, is_inferred, type,
from_id, to_id, to_id_supertype, to_id_sub_type, to_id_num)mc3_graph <- tbl_graph(
nodes = mc3_nodes_final,
edges = mc3_edges_final,
directed = TRUE
)str(mc3_graph)Classes 'tbl_graph', 'igraph' hidden list of 10
$ : num 1159
$ : logi TRUE
$ : num [1:3226] 0 0 0 0 0 0 0 1 1 1 ...
$ : num [1:3226] 1137 356 746 894 875 ...
$ : NULL
$ : NULL
$ : NULL
$ : NULL
$ :List of 4
..$ : num [1:3] 1 0 1
..$ : Named list()
..$ :List of 28
.. ..$ type : chr [1:1159] "Entity" "Entity" "Entity" "Entity" ...
.. ..$ label : chr [1:1159] "Sam" "Kelly" "Nadia Conti" "Elise" ...
.. ..$ name : chr [1:1159] "Sam" "Kelly" "Nadia Conti" "Elise" ...
.. ..$ sub_type : chr [1:1159] "Person" "Person" "Person" "Person" ...
.. ..$ id : chr [1:1159] "Sam" "Kelly" "Nadia Conti" "Elise" ...
.. ..$ timestamp : chr [1:1159] NA NA NA NA ...
.. ..$ monitoring_type : chr [1:1159] NA NA NA NA ...
.. ..$ findings : chr [1:1159] NA NA NA NA ...
.. ..$ content : chr [1:1159] NA NA NA NA ...
.. ..$ assessment_type : chr [1:1159] NA NA NA NA ...
.. ..$ results : chr [1:1159] NA NA NA NA ...
.. ..$ movement_type : chr [1:1159] NA NA NA NA ...
.. ..$ destination : chr [1:1159] NA NA NA NA ...
.. ..$ enforcement_type : chr [1:1159] NA NA NA NA ...
.. ..$ outcome : chr [1:1159] NA NA NA NA ...
.. ..$ activity_type : chr [1:1159] NA NA NA NA ...
.. ..$ participants : int [1:1159] NA NA NA NA NA NA NA NA NA NA ...
.. ..$ reference : chr [1:1159] NA NA NA NA ...
.. ..$ permission_type : chr [1:1159] NA NA NA NA ...
.. ..$ start_date : chr [1:1159] NA NA NA NA ...
.. ..$ end_date : chr [1:1159] NA NA NA NA ...
.. ..$ report_type : chr [1:1159] NA NA NA NA ...
.. ..$ submission_date : chr [1:1159] NA NA NA NA ...
.. ..$ jurisdiction_type: chr [1:1159] NA NA NA NA ...
.. ..$ authority_level : chr [1:1159] NA NA NA NA ...
.. ..$ coordination_type: chr [1:1159] NA NA NA NA ...
.. ..$ operational_role : chr [1:1159] NA NA NA NA ...
.. ..$ new_index : int [1:1159] 1 2 3 4 5 6 7 8 9 10 ...
..$ :List of 8
.. ..$ id : chr [1:3226] "2" "3" "5" "3013" ...
.. ..$ is_inferred : logi [1:3226] TRUE FALSE TRUE TRUE TRUE TRUE ...
.. ..$ type : chr [1:3226] NA "sent" NA NA ...
.. ..$ from_id : chr [1:3226] "Sam" "Sam" "Sam" "Sam" ...
.. ..$ to_id : chr [1:3226] "Relationship_Suspicious_217" "Event_Communication_370" "Event_Assessment_600" "Relationship_Colleagues_430" ...
.. ..$ to_id_supertype: chr [1:3226] "Relationship" "Event" "Event" "Relationship" ...
.. ..$ to_id_sub_type : chr [1:3226] "Suspicious" "Communication" "Assessment" "Colleagues" ...
.. ..$ to_id_num : chr [1:3226] "217" "370" "600" "430" ...
$ :<environment: 0x0000015fbacb8900>
- attr(*, "active")= chr "nodes"VisNetwork provides the user to understand relationships through interactivity. For instance:
The individual nodes can be selected from the drop-down menu to view its connected nodes and edges.
The hover tooltip provides additional details from fields such as content, coordination_type, findings, destination, operational_role, results, and jurisdiction_type based on the related id information from mc3_nodes_final.
# ---- 1. Define styles and legends ----
event_subtypes <- c(
"Communication", "Monitoring", "VesselMovement", "Assessment",
"Collaborate", "Endorsement", "TourActivity", "TransponderPing",
"Harbor Report", "Fishing", "Criticize"
)
relationship_subtypes <- c(
"Coordinates", "AccessPermission", "Operates", "Colleagues",
"Suspicious", "Reports", "Jurisdiction", "Unfriendly", "Friends"
)
node_legend_colors_plot <- c(
"Person" = "#88CCEE",
"Vessel" = "#D55E00",
"Organization" = "#117733",
"Location" = "#AA4499",
"Group"= "#CC79A7",
"Event" = "#DDCC77", # type level
"Relationship" = "#AF8DC3" # type level
)
node_legend_shapes_plot <- c(
"Person" = "dot",
"Vessel" = "triangle",
"Organization" = "square",
"Location" = "diamond",
"Group" = "circle plus",
"Event" = "star", # type level
"Relationship" = "square x" # type level
)
STYLES <- list(
node_label_dark = "black",
font_family = "Roboto Condensed"
)
# ---- 2. Prepare nodes ----
nodes <- mc3_nodes_final %>%
mutate(
label = ifelse(is.na(name), id, name),
# These parts are for pulling the related data from other fields
tooltip_extra = case_when(
type == "Event" & sub_type == "Communication" ~ content,
type == "Event" & sub_type == "Monitoring" ~ findings,
type == "Event" & sub_type == "VesselMovement" ~ destination,
type == "Event" & sub_type == "Assessment" ~ results,
type == "Relationship" & sub_type == "Coordinates" ~ coordination_type,
type == "Relationship" & sub_type == "Operates" ~ operational_role,
type == "Relationship" & sub_type == "Jurisdiction" ~ jurisdiction_type,
TRUE ~ NA_character_
),
title = paste0(
"<b>", label, "</b><br>",
"Type: ", type, "<br>",
"Sub-type: ", sub_type, "<br>",
ifelse(!is.na(tooltip_extra), paste0("<br><b>Details:</b> ", tooltip_extra), "")
),
# Fallback logic: if sub_type is NA or not in styling list, use type instead
group = ifelse(sub_type %in% names(node_legend_colors_plot), sub_type, type)
) %>%
select(id, label, group, title) %>%
distinct()
# ---- 3. Prepare directed edges (type == "sent") ----
edges <- mc3_edges_final %>%
filter(from_id %in% nodes$id & to_id %in% nodes$id) %>%
select(from = from_id, to = to_id)
# ---- 4. Build visNetwork ----
net <- visNetwork(nodes, edges, width = "100%", height = "600px") %>%
visEdges(arrows = list(to = list(enabled = TRUE, scaleFactor = 1.5))) %>%
visOptions(highlightNearest = TRUE, nodesIdSelection = TRUE) %>%
visIgraphLayout(layout = "layout_with_fr") %>%
visNodes(font = list(
size = 14,
color = STYLES$node_label_dark,
face = STYLES$font_family,
vadjust = -15
))
# ---- 5. Apply shape and color per group ----
for (group_name in names(node_legend_colors_plot)) {
net <- net %>% visGroups(
groupname = group_name,
color = node_legend_colors_plot[[group_name]],
shape = node_legend_shapes_plot[[group_name]]
)
}
# ---- 6. Add legend ----
used_groups <- unique(nodes$group)
legend_df <- tibble::tibble(
label = used_groups,
shape = node_legend_shapes_plot[used_groups],
color = node_legend_colors_plot[used_groups]
) %>%
distinct(label, .keep_all = TRUE) # remove duplicates just in case
net <- net %>% visLegend(
addNodes = legend_df,
ncol = 2, # number of columns
position = "left",
main = "Entity (Sub)Types", # title
useGroups = FALSE # show custom legend entries
)
# ---- 7. Render ----
netCheck if mapping is correct for type and sub_type.
mc3_nodes_cleaned %>%
group_by(type, sub_type) %>%
summarize(count = n()) %>%
arrange(-count) %>%
kable()| type | sub_type | count |
|---|---|---|
| Event | Communication | 584 |
| Relationship | Coordinates | 74 |
| Event | Monitoring | 70 |
| Relationship | AccessPermission | 68 |
| Event | VesselMovement | 46 |
| Relationship | Operates | 40 |
| Event | Assessment | 36 |
| Relationship | Colleagues | 30 |
| Entity | Location | 29 |
| Relationship | Suspicious | 28 |
| Event | Collaborate | 25 |
| Relationship | Reports | 25 |
| Event | Enforcement | 21 |
| Entity | Person | 18 |
| Entity | Vessel | 15 |
| Event | TourActivity | 13 |
| Relationship | Jurisdiction | 13 |
| Entity | Group | 5 |
| Entity | Organization | 5 |
| Relationship | Unfriendly | 5 |
| Event | TransponderPing | 3 |
| Event | HarborReport | 2 |
| Relationship | Friends | 2 |
| Event | Criticize | 1 |
| Event | Fishing | 1 |
# Split the 'from_id' column
mc3_edges_cleaned <- mc3_edges_cleaned %>%
separate(from_id, into = c("from_id_supertype", "from_id_sub_type", "from_id_id"), sep = "_", remove = FALSE, extra = "drop")
# Split the 'target' column into
mc3_edges_cleaned <- mc3_edges_cleaned %>%
separate(to_id, into = c("to_id_supertype", "to_id_sub_type","to_id_id"), sep = "_", remove = FALSE, extra = "drop")
# Find the number of unique types in each column
unique_counts <- mc3_edges_cleaned %>%
summarise_all(n_distinct) %>%
pivot_longer(cols = everything(), names_to = "column", values_to = "unique_count")
# Print the unique counts for each column
print(unique_counts)# A tibble: 12 × 2
column unique_count
<chr> <int>
1 id 3170
2 is_inferred 2
3 from_id 1052
4 from_id_supertype 48
5 from_id_sub_type 20
6 from_id_id 822
7 to_id 1156
8 to_id_supertype 71
9 to_id_sub_type 21
10 to_id_id 860
11 to_id_num 860
12 type 4# Check the mapping
mc3_edges_cleaned %>%
group_by(from_id_supertype, from_id_sub_type) %>%
summarize(count = n()) %>%
arrange(-count) %>%
kable()| from_id_supertype | from_id_sub_type | count |
|---|---|---|
| Event | Communication | 1620 |
| Green Guardians | NA | 90 |
| Relationship | Coordinates | 81 |
| Mako | NA | 78 |
| Reef Guardian | NA | 72 |
| Relationship | AccessPermission | 63 |
| Oceanus City Council | NA | 62 |
| Remora | NA | 62 |
| Event | Monitoring | 59 |
| Neptune | NA | 57 |
| EcoVigil | NA | 51 |
| Sentinel | NA | 47 |
| Davis | NA | 46 |
| The Lookout | NA | 44 |
| Event | VesselMovement | 43 |
| Relationship | Operates | 41 |
| The Intern | NA | 37 |
| Event | Assessment | 33 |
| Horizon | NA | 32 |
| V. Miesel Shipping | NA | 32 |
| Himark Harbor | NA | 31 |
| Mrs. Money | NA | 31 |
| Miranda Jordan | NA | 30 |
| Clepper Jensen | NA | 27 |
| Liam Thorne | NA | 26 |
| Relationship | Suspicious | 26 |
| Relationship | Reports | 25 |
| Rodriguez | NA | 25 |
| Event | Enforcement | 23 |
| Paackland Harbor | NA | 22 |
| Boss | NA | 21 |
| The Middleman | NA | 19 |
| Event | Collaborate | 17 |
| Nadia Conti | NA | 17 |
| Small Fry | NA | 17 |
| Marlin | NA | 15 |
| Osprey | NA | 15 |
| Samantha Blake | NA | 15 |
| Serenity | NA | 15 |
| Event | TourActivity | 14 |
| Relationship | Jurisdiction | 14 |
| Defender | NA | 12 |
| Knowles | NA | 12 |
| Seawatch | NA | 12 |
| Elise | NA | 10 |
| Glitters Team | NA | 9 |
| Haacklee Harbor | NA | 9 |
| Sailor Shifts Team | NA | 9 |
| Northern Light | NA | 8 |
| The Accountant | NA | 8 |
| Sam | NA | 7 |
| Relationship | Unfriendly | 5 |
| Conservation Vessels | NA | 4 |
| Kelly | NA | 4 |
| Event | HarborReport | 3 |
| Event | TransponderPing | 3 |
| Mariner’s Dream | NA | 3 |
| Recreational Fishing Boats | NA | 3 |
| Port Security | NA | 2 |
| Relationship | Colleagues | 2 |
| Tourists | NA | 2 |
| Diving Tour Operators | NA | 1 |
| Event | Criticize | 1 |
| Event | Fishing | 1 |
| Sailor Shift | NA | 1 |
# Check the mapping
mc3_edges_cleaned %>%
group_by(to_id_supertype, to_id_sub_type) %>%
summarize(count = n()) %>%
arrange(-count) %>%
kable()| to_id_supertype | to_id_sub_type | count |
|---|---|---|
| Event | Communication | 584 |
| Event | Monitoring | 240 |
| Relationship | AccessPermission | 225 |
| Relationship | Coordinates | 207 |
| Event | VesselMovement | 160 |
| Relationship | Colleagues | 160 |
| Relationship | Operates | 129 |
| Event | Assessment | 103 |
| Nemo Reef | NA | 102 |
| Mako | NA | 89 |
| Relationship | Reports | 72 |
| Oceanus City Council | NA | 68 |
| Relationship | Suspicious | 68 |
| Relationship | Jurisdiction | 63 |
| Event | Collaborate | 62 |
| Remora | NA | 60 |
| Event | Enforcement | 47 |
| Neptune | NA | 45 |
| Himark Harbor | NA | 38 |
| Reef Guardian | NA | 38 |
| Green Guardians | NA | 37 |
| Sentinel | NA | 36 |
| V. Miesel Shipping | NA | 35 |
| Event | TourActivity | 32 |
| Horizon | NA | 32 |
| Paackland Harbor | NA | 29 |
| Mrs. Money | NA | 26 |
| Boss | NA | 24 |
| EcoVigil | NA | 21 |
| Miranda Jordan | NA | 20 |
| Nadia Conti | NA | 20 |
| Protected areas | NA | 20 |
| Clepper Jensen | NA | 19 |
| Davis | NA | 18 |
| The Intern | NA | 18 |
| Liam Thorne | NA | 17 |
| Seawatch | NA | 16 |
| Event | TransponderPing | 12 |
| Sailor Shifts Team | NA | 12 |
| Relationship | Unfriendly | 11 |
| Serenity | NA | 11 |
| Marlin | NA | 10 |
| Restricted Zone | NA | 10 |
| Sam | NA | 10 |
| Restricted areas | NA | 9 |
| Rodriguez | NA | 9 |
| South Dock | NA | 9 |
| The Lookout | NA | 9 |
| Haacklee Harbor | NA | 7 |
| Samantha Blake | NA | 7 |
| The Middleman | NA | 7 |
| Eastern reefs | NA | 6 |
| Elise | NA | 6 |
| Knowles | NA | 6 |
| Northern quadrant | NA | 6 |
| Relationship | Friends | 6 |
| Western quadrant | NA | 6 |
| Dolphin Bay | NA | 5 |
| Eastern Boundary | NA | 5 |
| Northern Light | NA | 5 |
| Western Boundary | NA | 5 |
| Eastern quadrant | NA | 4 |
| City Officials | NA | 3 |
| Coral Point | NA | 3 |
| Defender | NA | 3 |
| E7 | NA | 3 |
| Event | HarborReport | 3 |
| Osprey | NA | 3 |
| Route C | NA | 3 |
| The Accountant | NA | 3 |
| Berth 14 | NA | 2 |
| Castaway Cove | NA | 2 |
| Eastern Islands | NA | 2 |
| Eastern Shoals | NA | 2 |
| Event | Criticize | 2 |
| Event | Fishing | 2 |
| Mariner’s Dream | NA | 2 |
| Southern Boundary | NA | 2 |
| Southern coastline | NA | 2 |
| Southern islands | NA | 2 |
| Azure Cove | NA | 1 |
| Conservation Vessels | NA | 1 |
| Eastern Coastline | NA | 1 |
| Glitters Team | NA | 1 |
| Kelly | NA | 1 |
| Port Security | NA | 1 |
| Recreational Fishing Boats | NA | 1 |
| Small Fry | NA | 1 |
| Southern quadrant | NA | 1 |
Under Event-Communication types: The edges target type and target subtypes matches the count of 584 for node to_id_supertype and node to_id_sub_type. However, there were only 581 count for content within the original node file. We then looked into duplicates.
# checking for duplicates
duplicate_values1 <- mc3_nodes_cleaned %>%
count(content) %>%
filter(n > 1)
# View duplicates
print(duplicate_values1)# A tibble: 5 × 2
content n
<chr> <int>
1 Boss, The Accountant here. Conservation vessels deploying underwater mi… 2
2 Davis here to V. Miesel Shipping. Crew reallocation from Remora to Nept… 2
3 Mrs. Money, this is Middleman. I've redirected Council's attention to o… 2
4 Rodriguez, Davis here. Maintain current position with Mako at Nemo Reef… 2
5 <NA> 575There were 4 duplicates within the content column. Upon checking the original data, one was the sender and the other was the receiver who received the same content. We left the data as it was.