Options
Aggregate binomial columns
Contributor II
Hello all,
I have a dataset that looks like:
User | Item
-------------
1 | Cheese
1 | Bread
2 | Milk
I'd like to mine the frequent item sets from this data. First thing I did was feed this to "Nominal to Binomial" which seems to work as expected, eg:
User | Cheese | Bread | Milk
------------------------------------------------------------
1 | true | false | false
1 | false | true | false
2 | false | false | true
What I now need to do is aggregate by user ID to generate:
User | Cheese | Bread | Milk
------------------------------------------------------------
1 | true | true | false
2 | false | false | true
I thought I could do this with the Aggregate operator, but that operator seems completely blind to the binomial columns; I can't find any way of selecting them.
What should I be doing here?
Thank you!
I have a dataset that looks like:
User | Item
-------------
1 | Cheese
1 | Bread
2 | Milk
I'd like to mine the frequent item sets from this data. First thing I did was feed this to "Nominal to Binomial" which seems to work as expected, eg:
User | Cheese | Bread | Milk
------------------------------------------------------------
1 | true | false | false
1 | false | true | false
2 | false | false | true
What I now need to do is aggregate by user ID to generate:
User | Cheese | Bread | Milk
------------------------------------------------------------
1 | true | true | false
2 | false | false | true
I thought I could do this with the Aggregate operator, but that operator seems completely blind to the binomial columns; I can't find any way of selecting them.
What should I be doing here?
Thank you!
Tagged:
0
Answers
(I think there's an FP Growth example somewhere in the Rapid Miner tutorials which shows this.)
Greets,
René
Thanks for the reply. This doesn't seem to do what I want; again, the binomial columns appear invisible to the Pivot operator.
The FPGrowth example has the data already in a collapsed format, so it can be passed directly to the FPGrowth operator. I can't seem to do this in RapidMiner.
However, I managed to hack around it by using GROUP_CONCAT in the SQL query, use the Split operator to break up the column into new attributes, then feed it to Nominal to Binomial and off to the FPGrowth operator!
Shame my data didn't seem to have any sets
Starting from your 2nd table, I'd first think of logcally OR'ing the observations for each user ID. I'm sure there are some nifty matrix operations for archieving that. But, as a beginner, I got to pass ...
I've encountered the same problem, any solutions? Aggregate operator is not working for binominal values.