How to sort association rules by lhs or rhs in R

Question

I have set of rules from apriori algorithm. Sorting them by lift, confidence or support is easy:

rules.sorted = sort(rules, by="lift")

But let's say I have small set of rules with a few different rhs elements. I would like to view rules sorted by those rhs elements (alphabetically). Is there a way to do that?

I.e from this:

    lhs       rhs support   confidence lift    
1   {A}    => {B} 0.3919252 0.9431280  1.930940
2   {B}    => {A} 0.3919252 0.8024194  1.930940
3   {E,C}  => {A} 0.3535204 0.7995546  1.924047
4   {F,I}  => {F} 0.3924175 0.9005650  1.868281
5   {H}    => {G} 0.4194978 0.9659864  1.864941
6   {C,D}  => {A} 0.3653373 0.7141482  1.718525

To this:

    lhs       rhs support   confidence lift    

2   {B}    => {A} 0.3919252 0.8024194  1.930940
3   {E,C}  => {A} 0.3535204 0.7995546  1.924047
6   {C,D}  => {A} 0.3653373 0.7141482  1.718525
1   {A}    => {B} 0.3919252 0.9431280  1.930940
4   {F,I}  => {F} 0.3924175 0.9005650  1.868281
5   {H}    => {G} 0.4194978 0.9659864  1.864941

Here is an example of rules that I work with:

library(arules)
download.file("http://archive.ics.uci.edu/ml/machine-learning-databases/mushroom/agaricus-lepiota.data", "mush.data");
dataset = read.table("mush.data", header = F, sep=",", na.strings= "*")
tr <- as(dataset, "transactions")
param  = new("APparameter", "confidence" = 0.9, "support" = 0.7, "minlen"= 2L, "maxlen" = 2L, "target" = "rules") 
rules <-apriori(tr,param)
dput(rules)

When I try to sort by rhs I get `Error in '[.data.frame'(q, , pmatch(by, colnames(q)), drop = FALSE) : undefined columns selected` — Ewa Janikowska, Nov 17 '16 at 14:27
According to `?sort` from arules, the method only works with `measures stored in the association's slot quality`, you might want to subset your rules instead if interested in specific items. — mtoto, Nov 17 '16 at 14:41
@Cath I edited my question to include `dput`. @mtoto I know I can subset with given rhs, but I would have to do this separately for each element. I'm looking for an easier way — Ewa Janikowska, Nov 17 '16 at 14:43
Class “transactions” object - `tr <- as(dataset, "transactions")` — Ewa Janikowska, Nov 17 '16 at 15:31
As you can see from the code above it's a data frame: `dataset = read.table("mush.data", header = F, sep=",", na.strings= "*")` — Ewa Janikowska, Nov 17 '16 at 15:44

score 4 · Answer 1 · answered Nov 17 '16 at 23:12

4

bergant's solution is exactly what you need. I would just directly reorder the rules in the rules object:

rules_sorted <- rules[order(labels(rhs(rules)))]

The new rules object is now sorted.

answered Nov 17 '16 at 23:12

Michael Hahsler

2,965
1
12
16

score 3 · Accepted Answer · answered Nov 17 '16 at 22:05

3

Extract basic information from your rules object to a data frame, like this:

rules_info <-
  data.frame(
    LHS = labels(lhs(rules)), 
    RHS = labels(rhs(rules)),          
    quality(rules)
  )

Then sort it like a normal data frame:

rules_info[ order(rules_info$RHS), ]

answered Nov 17 '16 at 22:05

bergant

7,122
1
20
24

Thanks, that's what I was looking for! – Ewa Janikowska Nov 17 '16 at 22:28
I want to do something like you suggested. But this code converts `rhs` and `lhs` to factors and the actual values appear as strings. I want `rhs` and `lhs` to appear as sets in `rules_info` data frame. Is it possible? – Muhammad Adeel Zahid Jan 03 '19 at 12:14
1

@MuhammadAdeelZahid use stringsAsFactors = FALSE when creating a data.frame – bergant Jan 07 '19 at 15:42

How to sort association rules by lhs or rhs in R

2 Answers2