I have this dataset which I really need to use association rules techniques on. The dataset has like 90 variables, many of which are ordinal. Thing is, the data is already coded using numbers instead of strings (e.g. bread = 4 instead of "bread") as well as some re-scaled numerics such as 1 = 1%-10%".

What I have so far:

from apyori import apriori
#Convert dataframe to list
val_list= []
for row in range(1,5530):
    val_list.append([str(data.values[row,column]) for column in range (0,90)])
    print('Row ', row, ' ok')

apr = apriori(val_list,min_support=0.1,min_confidence=0.2,min_lift=2)    
result = list(apr)

Still, this way I don't get the feature names in the frequent "baskets" so it's not much use, since I have baskets like [33, 1, 8, 8, 1, 1] with no idea what the numbers might be referring to. What can I do and/or how do I prepare the data for association rule mining?


1 Answer 1


Create a dictionary that contains the coded variables as keys and the item names as values.

So it would look like:

dicty = {4: "bread", 7: "milk", 9: "toothpaste"}

Constructing dictionaries in python is really easy if you have them in a table or excel spreadsheet.

dicty = {i:j for i,j in zip(coded_list,normal_list)}

where coded_list is the list of variables in numbers, and normal_list is the list of variables in their categorical names.

Once you have a dictionary you can simply convert them like this:

name = dicty[9]

and it should return toothpaste for name.

  • $\begingroup$ Not that easy though. I have 90 features that are different from each other. Am I supposed to do that for ALL of them? Could I somehow make it so the code is "Feature_Name+Coded_Number"? For example code "9" for a feature named "ABC" would be "ABC9". Like how factor variables appear in regressions in R. $\endgroup$
    – NeR0
    Commented Jun 23, 2020 at 20:25
  • $\begingroup$ @NeR0 yes, you are supposed to do it with all of them. You want the feature name because that's how you understand the data. The computer wants the coded data because that's how it understands the data. IA dictionary allows you to interconvert between the two. $\endgroup$ Commented Jun 24, 2020 at 15:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.