Hyperparameter Tuning // Grid Search Always Picking Biggest Numbers

emrek
emrek Registered Posts: 4

I am using Random Forest algorithm. Also, I am trying tune hyperparameters by using Grid Search. Grid Search picking biggest number hyperparameters all the time. I am sharing a pictures about my problem. Is there any reason of this problem?


Picture 1.pngPicture 2.png

Answers

  • Miguel Angel
    Miguel Angel Dataiker, Dataiku DSS Core Designer, Dataiku DSS ML Practitioner, Dataiku DSS Adv Designer, Registered Posts: 118 Dataiker

    Hi,

    There is not a problem here as far as we can see.

    In training the model, it was found the 'biggest numbers' provide better results. This is also mentioned in the uploaded screenshot:'Higher values generally increase the quality of the prediction'.

  • Rickh008
    Rickh008 Dataiku DSS Core Designer, Registered Posts: 15 ✭✭✭✭

    Have you tested against the model using validation/hold-out dataset?

    A more generalized model may perform better with new data. It's unsurprising that a more advanced model performs better on training/test data.

  • emrek
    emrek Registered Posts: 4

    Hi, yes i did. I was thinking that biggest number always can't be best one for model. I am sharing my project flow and hyperparameter optimization for random forest model. We can clearly see line is always going up on hyperparameter optimization plot.

Setup Info
    Tags
      Help me…