Machine Learning Studio- What is the correct way to apply feature selection? RRS feed

  • Question

  • I recently ran an experiment on Machine Learning Studio with Fisher LDA module but I am not sure what is the correct way to split the training and testing data.

    At first I applied the feature selection module and then split the transformed dataset into two sets for training and testing. The accuracy was 100% which is too high for me to believe. Then I tried another way which is to split the raw data before applying the module. This time the accuracy was only 60%

    I am a beginner in this area and I am confused which one is the correct way. Can somebody explained what's going wrong here?

    Thank you!!

    Thursday, January 31, 2019 3:37 AM

All replies

  • Hi Kelvin,

    Thanks for your feedback. Can you please share the experiment you are working on so that we can know your scenario more? 



    Thursday, January 31, 2019 6:37 PM