"Market-Basket-Wrong-Approach?"

SunnyLotusFlowe

Hello everyone,

i wanted to use some concept hierarchy for a Data set like this

C-ID | Product
001 |  apple
002 | Banana
001 | Banana
003 | Chair

_____________________
i want to map all apple and Banana Values to Fruit and insert new transaction-examples

C-ID | Product
001 |  apple
002 | Banana
001 | Banana
003 | Chair
001 |  Fruit
002 | Fruit
001 |  Fruit

Is this approach wrong? does the support vaules for the rules will stay the same ?

After the pivoting of this i get the following:

C-ID | apple | Banana | Fruit | chair
001 |    1    |      1    |  2    |    0
002 |    0    |      1    |  1    |    0
003 |    0    |      0    |  0    |    1

is it a problem that the third column has a '2'  ??? if i apply now the numerical to binominal i will get lost information i think
because the '2' will get mapped to true.

My two questtions are:
1. if i mine the table above, will i get something that is wrong ??? because of the value '2'
2. Is my approach for the concept hierarchy the right one ???

greetings

Lotus

SunnyLotusFlowe

is it to complicated what i have written ?

haddock

Maybe.

SunnyLotusFlowe

ok then in clear words:


do i 'deform' the data, by adding further Customer ID to it ?
i think i change the support of the rules, which i extract later...

haddock

Maybe.

SunnyLotusFlowe

then is there another 'trick' to build Concept Hierarchies without 'deforming' the data?

are there any standart-solutions ?

haddock

All the ones I know of involve spending money, are you OK with that?

SunnyLotusFlowe

oh, my interests of RapidMiner are non-commercial. i just do some kind of 'report' and wanted to know what methods are implemented.

but if i understand u correctly theres no possible simple trick with the operators avaible ... ?

haddock

but if i understand u correctly theres no possible simple trick with the operators avaible ... ?

You did not understand me correctly.

IngoRM

Hey Lotus,

My two questtions are: 1. if i mine the table above, will i get something that is wrong Huh because of the value '2' 2. Is my approach for the concept hierarchy the right one Huh

what do you expect by us? To read all the papers and provide you with condensed knowledge about this topic for every single question you got? Sorry, but I really like Hadocks answers here: you do some kind of report (you have to do?) and you are not willing to dive deep enough into the theoretical details or into the software you want to report about? Instead of that you totally rely on the answers of the community - which might be correct. Or wrong. Maybe.

All I want to add is: if you are clear about the idea of what you want to achieve (in data analysis) with RapidMiner I am giving you my word that you can achieve it with the available operators. I would place a bet here: tell me your analysis problem and I am betting that I will be able to create the process for this.

Cheers,
Ingo

SunnyLotusFlowe

what do you expect by us? To read all the papers and provide you with condensed knowledge about this topic for every single question you got? Sorry, but I really like Hadocks answers here: you do some kind of report (you have to do?) and you are not willing to dive deep enough into the theoretical details or into the software you want to report about? Instead of that you totally rely on the answers of the community - which might be correct. Or wrong. Maybe.
________________________________________________________________________
well i try my best to understand the concepts, but if i dont know what to do i ask here in the forum. and if some nice person answer me i try his idea and if it work for my purpose i write it down. this is btw one of the last questions.

i think about it on my own, but i got confused from time to time and therefore asking.
___________________________________________________________________________________


in this topic i just wanted to see if i got the right idea to set up a Concept Hierarchy in basket market analysis. but i got confused if i crush the context of the underlying data with my idea.  i can extract some assoc rules with my 'transformation' (= adding generalized items) but i'm not sure 100% that values like the support are correct.


greetings

Lotus

haddock

Whatever you thought you were doing, it hasn't been working brilliantly with the people you were trying to get answers from. Because you didn't get an answer within two hours you got arsey, but you say...

and if some nice person answer me i try his idea and if it work for my purpose i write it down.

Can you not understand it might look different to others?

SunnyLotusFlowe

if there was something that gets you angry or something like that, i have to apologize for that. i never wanted to get someone angry (or something like that) I hope noone is offended by something i wrote....
_____________________________________________________________________

but i build a workflow (to answer 1 of my questions):
i used 1 dataset like

cardid,Product
001,apple
001,banana
002,banana

and the other

cardid,Product
001,apple
001,banana
002,banana
001,fruit
002,fruit

to check the values like support | confidence ....
In both result there were the following assoc rules:

apple -> banana
banana -> apple
Both rules got the same values in the results. I mean using the first data set the rule 'apple -> banana' got the same values than with the other data set and vice versa.

Therefore i conclude that inserting generalized transaction-examples wont have an effect on the values of assoc rules that are not containing the new inserted items...


again sry for the trouble...

greetings

Lotus

IngoRM

Hi,

no need to be sorry. I think the bottom line here is that the way you are asking combined with the amount of questions will not really motivate people.

Back to the topic: I would expect that adding those artificial transactions would actually change the support values. Look at your example: For the first data set the support of "apple" is 1/3 and for the second it is 1/5. So the question arises why you don't analyse both problems separately? First time based on the items and another time based on their groups. In this case you would however not get rules mixing items and groups. If this is necessary, other actions - like adding those artificial transactions - might be the best you can do.

Cheers,
Ingo