CODEWITHSUNDEEP
ralgorithmassociationsapriori
rebe1022
rebe1022

Reputation: 3

Apriori algorithm in R, not negative rules

I have a large binary data set where I wish to run an apriori algorithm in R. The problem is at the algorithm is making rules of all the 0's, where I only wish to look at the 1's. As for example get these rules: 

        lhs                                                        rhs     support   confidence lift      count
[1]    {SPA=0,SPD=0,SPE=0,SPF=1,SPJ=0}                         => {SPC=0} 0.2036065 0.9866727  1.0174854  6515
[2]    {SPA=0,SPD=0,SPE=0,SPF=1}                               => {SPC=0} 0.2163885 0.9864653  1.0172715  6924
[3]    {SPA=0,SPD=0,SPF=1,SPJ=0}                               => {SPC=0} 0.2070754 0.9852788  1.0160479  6626

Does anyone know how to only look for the rules where the variables are 1 and not 0? Thank you!

Upvotes: 0

Views: 157

Answers (2)

Michael Hahsler
Michael Hahsler

Reputation: 3075

The easiest way to fix this is to make the matrix logical before you create the transactions. For matrix m you can do the following:

storage.mode(m) <- "logical"
trans <- as(m, transactions)

Upvotes: 0

G5W
G5W

Reputation: 37641

You can control this using the appearance argument to apriori. Since you do not provide data, I will use the built-in Adult data as an example, but I think that you need to add appearance=list(rhs = "SPC=1") to your apriori statement.

Example

I will generate only rules for which the rhs is native-country=United-States

rules <- apriori(Adult, 
    parameter = list(supp = 0.4, conf = 0.6, 
        minlen=2, target = "rules"),
    appearance=list(rhs = "native-country=United-States")
)

inspect(rhs(rules[1:5]))
    items                         
[1] {native-country=United-States}
[2] {native-country=United-States}
[3] {native-country=United-States}
[4] {native-country=United-States}
[5] {native-country=United-States}

Addition

I thought that you only wanted SPC=1 on the rhs. Based on your comments, I now think that you want to generate rules that contain no XYZ=0 items at all. You can also get this with appearance. First identify the possible items with XYZ=0, then use appearance to exclude these. I do not know what your variables are called, so I am calling the transactions TransactionData

## identify items to exclude
excluded <- grep("=0", itemLabels(TransactionData), value = TRUE)

Then add this to your apriori statement.

appearance=list(none = excluded)

Upvotes: 0

Related Questions