for loop to determine mutually exclusive/exhaustive network membership - r

Question

I want to run a for loop that assigns group_ids to a dataset that have two columns, name and location. The goal is to plot large mutually exclusive and exhaustive networks. it is sort of like contract tracing, where individuals interact with locations and then, as a function of both individual interaction and locations, you get a finite network of people who belong to a group. So A and C both interact with location B. Both A and C are in the same group, but so is every other person who interacted with location B. Likewise, any other locations, x1-xn, that interacted with A or C or anyone else from location B is part of the same group. And so forth.
Basically, in the for loop, I want to do the following. First, for the first row in the dataframe, I want to check if the name exists in the name_group_id list initialized. If it does, then assign the group_id associated with that name. If not, then check if the location exists in the location_group_id list. If it does, return that group_id, if it does not, then create a new group_id and add this row's name/group_id to the name_group_id and the location/group_id to the location_group_id
This code does not work yet, as row three should have group_id 1 because name interacted with location 1 in row 1. But it seems close. Help tweaking this code would be awesome.
Thanks, everyone.
# Create data frame with the sample data
df <- data.frame(name = c("a", "a", "b", "b", "b", "c", "c", "d", "d", "e", "e", "f", "g", "g", "h"),
location = c(1, 2, 1, 3, 4, 3, 2, 5, 6, 7, 8, 4, 9, 10, 5))
# Initialize two lists
name_id <- list()
location_id <- list()
# Counter for creating new group IDs
group_id_counter <- 0
# For loop that implements the logic
for (i in 1:nrow(df)) {
row <- df[i, ]
name <- row$_name
location <- row$location
# Check if the name exists in name_id
if (name %in% names(name_id)) {
group_id <- name_group_id[[name]]
} else {
# Check if the location exists in location_group_id
if (location %in% names(location _group_id)) {
group_id <- location _group_id[[location ]]
} else {
group_id_counter <- group_id_counter + 1
group_id <- paste0("Group ", group_id_counter)
name_id[[name]] <- group_id
location _group_id[[location ]] <- group_id
}
}
# Assign the group_id to the current row
df[i, "group_id"] <- group_id
}
Right now, this code would return group_id 2 to row three, but it should be group_id 1 because location 1 was already seen in row 1 and assigned group_id 1. Help would be greatly appreciated.
Expected results for the fake data frame supplied are:
name location Group ID
a 1 1
a 2 1
b 1 1
b 3 1
b 4 1
c 3 1
c 2 1
d 5 2
d 6 2
e 7 3
e 8 3
f 4 1
g 9 4
g 10 4
h 5 2

Answer 1

As Gregor Thomas mentioned this is the same as determining connected components in graphs. Another term for this is equivalence groups. The code below is a bit of hack using an internal function of one of my packages (on CRAN); you could do this using the external functions but then you have to convert the data into the right form; this is easier:
First make a list of all node 'identifiers':
x <- c(unique(df$name), unique(df$location))
Determine which nodes belong to the same group
res <- reclin2:::equivalence(x = x,
df$name, df$location)
The result res is a vector with identifiers for each group for each value in x. To get the end result:
res[match(df$name, x)]
Edit: the OP mentioned that the datasets are large; this should work on large datasets.