Stream: helpdesk (published)

Topic: AUC/ROC with Lighthouse.jl

view this post on Zulip QuBit (Jul 18 2021 at 23:46):

Hello Everyone:

I'm working with @load_boston from the MLJ
library. How would one of you apply the
Lighthouse.jl package to generate an ROC
curve?

Thank you,

view this post on Zulip Eric Hanson (Jul 19 2021 at 08:14):

What have you tried? If you include some code showing what you tried it will help (usually easier to fix it than start from scratch)

view this post on Zulip QuBit (Jul 19 2021 at 23:35):

Hello Folks:

I am attempting to implement instructions in the
Lighthouse.jl package to generate a AUC/ROC curve:

The data I am using has the structure:

TransportDist= ["Far", "Average", "Close"]
ground_ratio =[17,23,44,16,12,15,16,12,14,29]
predicted_ratio[13,15,16,7,20,12,14,16,18, 24]

Then applying the confusion_matrix method as:

confusion = Lighthouse.confusion_matrix(length(TransportDist), zip(predicted_ratio, ground_ratio))

After, I am plotting this fig using: 

fig, ax, p = plot_confusion_matrix(confusion, Classes, :Row)

At this stage, I am wondering how I could assign categories
to the class object from a DataFrame?

Also, could I feed in a DataFrame instead of an array for the
ground_ratio and predicted_ratio objects?

Last updated: Oct 02 2023 at 04:34 UTC