NetworkD3 sankey chart in R creating really messy links, even though my code should be good. What is wrong?

Question

I am trying to create a sankey diagram but the connections turn out all messed up. Here is my data:

#> A tibble: 61 x 3
#> # Groups:   id_2 [55]
#>     id_2 origin                target                     
#>    <int> <fct>                 <chr>                      
#>  1     4 Pendlerkort i app     Rejsekort                  
#>  2    38 Ungdomskort           Rejsekort                  
#>  3    84 Rejsekort             Rejsekort                  
#>  4    89 Rejsekort             Pendlerkort i app          
#>  5    95 Rejsekort             Rejsekort                  
#>  6   112 Rejsekort Pendlerkort Rejsekort                  
#>  7   118 Pendler20             Rejsekort med pendler kombi
#>  8   140 Pendlerkort i app     Rejsekort                  
#>  9   167 DSB Orange            Pendler20                  
#> 10   193 Rejsekort             Rejsekort                  
#> # ... with 51 more rows

sankey_plot <- 
  tibble::tribble(
    ~id_2,  ~origin,                  ~target,
    4L,     "Pendlerkort i app",      "Rejsekort",
    38L,    "Ungdomskort",            "Rejsekort",
    84L,    "Rejsekort",              "Rejsekort",
    89L,    "Rejsekort",              "Pendlerkort i app",
    95L,    "Rejsekort",              "Rejsekort",
    112L,   "Rejsekort Pendlerkort",  "Rejsekort"
  )

Then I run the following code:

# create nodes
nodes <- data.frame(
  name=c(as.character(sankey_plot$origin),
         as.character(sankey_plot$target)) %>% unique()
)

# Get values  
sankey_plot <- sankey_plot %>% group_by(origin, target) %>% 
  summarise(values = n())

sankey_plot <- sankey_plot %>% 
  dplyr::ungroup()

# Create source and target ID's
sankey_plot$IDsource <- match(sankey_plot$origin, nodes$name)-1 
sankey_plot$IDtarget <- match(sankey_plot$target, nodes$name)-1

# Plot
sankeyNetwork(Links = sankey_plot, Nodes = nodes,
              Source = "IDsource", Target = "IDtarget",
              Value = "values",
              NodeID = "name",
              # fontSize = 10 
              nodeWidth = 20,
              sinksRight = TRUE)

This gives me a messy plot looking like this:

I have no idea what I am doing wrong. I have tried to create the nodes in a few different ways. Also where I don't group_by() and simply let each row represent each movement (instead of having a count). The plot ends up looking the same

Answer 1

Given the sample data you provided, the loops are expected. For instance, line 3 represents a link that starts and ends at the same node, and line 1 represents a link from a node ("Pendlerkort I app") that follows the "Rejsekort" node and then links back to the "Rejsekort" node again.

sankey_plot
#> # A tibble: 5 × 5
#>   origin                target            values IDsource IDtarget
#>   <fct>                 <chr>              <int>    <dbl>    <dbl>
#> 1 Pendlerkort i app     Rejsekort              1        0        2
#> 2 Rejsekort             Pendlerkort i app      1        2        0
#> 3 Rejsekort             Rejsekort              2        2        2
#> 4 Rejsekort Pendlerkort Rejsekort              1        3        2
#> 5 Ungdomskort           Rejsekort              1        1        2

If what you're expecting is that there are different nodes that all have the same name "Rejsekort", then you will have to distinguish those nodes in your data.

library(dplyr)
library(networkD3)


sankey_plot <- 
  tibble::tribble(
  ~id_2,  ~origin,                  ~target,
  4L,     "Pendlerkort i app",      "Rejsekort",
  38L,    "Ungdomskort",            "Rejsekort",
  84L,    "Rejsekort",              "Rejsekort",
  89L,    "Rejsekort",              "Pendlerkort i app",
  95L,    "Rejsekort",              "Rejsekort",
  112L,   "Rejsekort Pendlerkort",  "Rejsekort"
)

dupe_targets <- sankey_plot$target == "Rejsekort"
sankey_plot$target[dupe_targets] <-
  paste0(sankey_plot$target[dupe_targets], "_", seq_along(sankey_plot$target[dupe_targets]))

sankey_plot$values <- 1

# create nodes
nodes <- data.frame(
  name=c(as.character(sankey_plot$origin),
         as.character(sankey_plot$target)) %>% unique()
)

# Create source and target ID's
sankey_plot$IDsource <- match(sankey_plot$origin, nodes$name)-1 
sankey_plot$IDtarget <- match(sankey_plot$target, nodes$name)-1


sankey_plot
#> # A tibble: 6 × 6
#>    id_2 origin                target            values IDsource IDtarget
#>   <int> <chr>                 <chr>              <dbl>    <dbl>    <dbl>
#> 1     4 Pendlerkort i app     Rejsekort_1            1        0        4
#> 2    38 Ungdomskort           Rejsekort_2            1        1        5
#> 3    84 Rejsekort             Rejsekort_3            1        2        6
#> 4    89 Rejsekort             Pendlerkort i app      1        2        0
#> 5    95 Rejsekort             Rejsekort_4            1        2        7
#> 6   112 Rejsekort Pendlerkort Rejsekort_5            1        3        8


sankeyNetwork(Links = sankey_plot, Nodes = nodes,
              Source = "IDsource", Target = "IDtarget",
              Value = "values",
              NodeID = "name",
              # fontSize = 10 
              nodeWidth = 20,
              sinksRight = TRUE)