KNN prediction performance

mehrdadmehrdad Member Posts: 3 Contributor I
edited November 2018 in Help

Hello All,

 

I am new here and I need your help. Actually, I have a training dataset and a test data set to be classified via kNN classification.  Training---->Knn--->apply model and Test --->apply model... I do not want use cross-validation or split validation but can someone tell me how to measure the performance of knn regarding the prediction of classifying my test dataset. I dont know how connected the output of apply model as an input the performance classification to know how good is the prediction of my classification.    

 

Tnx

Tagged:

Answers

  • landland RapidMiner Certified Analyst, RapidMiner Certified Expert, Member Posts: 2,525   Unicorn

    Hi,

    RapidMiner contains some very useful tutorials and explanations for beginners. It's a bit like the tutorials you have in games nowadays to explain what to do. I would recommend to check them out: Click on new process, select "Learn" and then start from the beginning to understand the principle workings and meanings of ports, parameters and colors, etc...

    The chapter 3 of Model, Scoring and Validation will exactly match your question. 

     

    Greetings

     Sebastian

    mehrdad
  • mehrdadmehrdad Member Posts: 3 Contributor I

    Thank Sebastian,

     

    I've checked some of them and I got results in the case of using kNN as sub process of cross or split validation but I have no idea how to evaluate  the score of the prediction of data which is shown in the attached pic.

     

    But-away tnx for your follow up

  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 2,113  RM Data Scientist

    Hi,

     

    the per port of cross validation is already delivering the performance. 

     

    otherwise- you can simply connect the lab port of Apply Model (2) with the lab port of Performance(2) and look at the result.

     

    ~Martin

    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
    mehrdad
  • mehrdadmehrdad Member Posts: 3 Contributor I

    Hi,

     

    Tnx for your reply. u right the cross-validation is already giving us the performance of classification but it is not the performance of predicting test set which is directly connected to the apply model 2. I wanna know that how good is my prediction regarding my classification approach. I wanted to connect out put of apply model 2 into performance directly but I got couple of errors.( it says u dont have label , or u need set criterion, but i dont know how!)

    If you want to know about the dependency of your classification performance with predicting of new test data set (retrieve 08 ), how do u measure it ?

     tnx in advance 

  • mschmitzmschmitz Administrator, Moderator, Employee, RapidMiner Certified Analyst, RapidMiner Certified Expert, University Professor Posts: 2,113  RM Data Scientist

    Ahhh, in order to apply Performance, the incoming data set needs to have a the label tagged. So you simply need to use a Set Role operator and set your label variable to label.

     

    ~Martin

    - Head of Data Science Services at RapidMiner -
    Dortmund, Germany
Sign In or Register to comment.