Due to recent updates, all users are required to create an Altair One account to login to the RapidMiner community. Click the Register button to create your account using the same email that you have previously used to login to the RapidMiner community. This will ensure that any previously created content will be synced to your Altair One account. Once you login, you will be asked to provide a username that identifies you to other Community users. Email us at Community with questions.

Logistic Regression - how confidence is computed

kloriklori Member Posts: 3 Contributor I
edited March 2020 in Help
Hi everyone!

I have a question regarding rapidminer and i hope somebody can help me please : )

I've created a Workflow based on the Weka-Logistic Regression. How the coefficents are calcuated and so on is not the problem.
Furthermore I built the Workflow so, in the end it created a csv file which applies the model on my input.csv.

Now I got the confidence of yes and no, which is the probability of course (I hope so).
I was little bit  skeptic so i computed the coefficients for one row of data and i got a prob which deviates from the computed prob by rapidminer.
I was wondering, if its possible or to track how Rapdiminer computes the confidence for the data?

With kind regards
Klori
Tagged:

Answers

  • earmijoearmijo Member Posts: 271 Unicorn
    Upload a dataset showing the problem.

    Confidences are calculated:

    p = exp( X*beta ) / ( 1 + exp(X*beta) ) ,   

    1-p = 1/ ( 1 + exp(X*beta) )

    The labels might be reversed ( the model might have been run for the 'No' class instead than the 'Yes' class.

  • kloriklori Member Posts: 3 Contributor I
    Hi earmijo : )

    Thanks for your help, but what I would like to know, if i can see how it calculates it?
    For example, i see which coeffiecents it takes to calc the p?

    One example, the coefficients are for yes:

    score_1 -> (null)
    score_2 -> -0,8628
    age-> -0,2192
    day -> -0,0423
    time -> 0,0923
    gender -> 0,3161 (binomial)
    segment -> 0,6009
    payment -> -0,078 (binomial)
    telefon -> 0,2288 (binomial)
    Constant -> -3,5047

    confidence(yes) -> 1,42%
    confidence(no) -> 98,58%

    I was testing around and set telefon to 0, gender to 0 and I come to 1,77% yes
    And if i take every coefficient it gives me 3,02%

    I would very much appreciate the way to see how rpaidminer takes the coefficients and compute the p.
    Thanks for your help earmijo

    Kind regards,
    Klori


    ps: Because i am using W-Lgistic i want to see the Weka result via rapidminer. Even thpugh it says "Creating weka_result, please wait..." it seems it doesnt load.
    Is this normal? Do i need to install something extra or so?
Sign In or Register to comment.