search

Laurae2 / Laurae

Advanced High Performance Data Science Toolbox for R by Laurae
204 stars 50 forks source link

daForest #11

Open ssobel opened 6 years ago

ssobel commented 6 years ago

Hello Laurae, this is is not an issue per se, but a question/suggestion... is there a way to create the function of daForest using your cascade & MG scanning algorithms? Thank you

daForest.pdf

Laurae2 commented 6 years ago

@ssobel You can use the Cascade Forest implementation with boosting enabled.

Then, predict using the layer parameter of the prediction function to span all the layers (so you have prediction for 1st layer, 1st/2nd layer together, 1st/2nd/3rd layer together, etc.).

Once done, take the combination resulting in the highest performance of the metric of your choice.

ssobel commented 6 years ago

Thank you! I will try that.