How to detect and remove recursive structure in hierarchical data in R?答案

【问题标题】：How to detect and remove recursive structure in hierarchical data in R?How to detect and remove recursive structure in hierarchical data in R?
【发布时间】：2022-12-01 23:19:32
【问题描述】：

here is an example data like mine. Think of it as a dataset of user activity on a website.

time	from	to
t0	A	a
t1	a	b
t2	a	c
t3	a	d
t4	b	x1
t5	b	x2
t6	c	y1
t7	x1	a

I want to generate hierarchical data. But like last row in dataset, there is infinite loop.

Parent is a then child is b.

Parent is b then child is x1.

Parent is x1 then child is a again. So here there is an infinite loop. So I can't use my original data because of that.

So my question is, how can i remove rows like the end?

Thank you.

I've tried to write some loop functions. But i couldn't.

PS: Edit example data

【问题讨论】：

One strategy for recognizing an infinite loop is to look for the new node in the list of nodes already handled. But without seeing your existing code, I can't tell you how to add this test.
I m not actually that good at nodes. Still learning. But i've tried code like that: (data.tree library) test=data.table(Parent = c('A','a','a','a','b','b','c','x1'), Child = c('a','b','c','d','x1','x2','y1', 'a')) data.tree::FromDataFrameNetwork(test) and I've got error for this code. Error is: "Error: C stack usage 7955072 is too close to the limit". I think this error occurs due to loop @user2554330
That error is typically a sign of an infinite recursion, e.g. f <- function() f() (but usually more complicated).

标签： r data.tree

【解决方案1】：

Your data frame represents the edgelist of a directed graph, and the operation you are describing is known asfinding the acyclic subgraphof your graph.

Using the igraph package, you can explicitly turn your data frame into a graph, find the edge(s) that cause an infinite loop using feedback_arc_set, subtract these from your graph, then convert the resulting directed acyclic graph back into a data frame:

library(igraph)

g <- graph_from_data_frame(df[c(2:3, 1)])
g <- g - feedback_arc_set(g)
new_df <- igraph::as_data_frame(g)[c(3, 1:2)]

new_df
#>   time from to
#> 1   t0    A  a
#> 2   t1    a  b
#> 3   t2    a  c
#> 4   t3    a  d
#> 5   t4    b x1
#> 6   t5    b x2
#> 7   t6    c y1

Note that this has correctly identified row t7 as causing a feedback arc and removed it from your data frame.

Data in reproducible format

df <- structure(list(time = c("t0", "t1", "t2", "t3", "t4", "t5", "t6", 
"t7"), from = c("A", "a", "a", "a", "b", "b", "c", "x1"), to = c("a", 
"b", "c", "d", "x1", "x2", "y1", "a")), class = "data.frame", row.names = c(NA, 
-8L))

【讨论】：

Wow! This is a very clever answer!