Stream: helpdesk (published)

Topic: AUC/ROC Curves With MLBase & Lighthouse

QuBit (Jul 19 2021 at 23:35):

@Eric Hanson Thank you.

I was caught up in the preprocessing steps.
The main ones I encountered that are
essential for working with Lighthouse are:

1. Formatting Floats to Integers using:

actual_test = parse.(Int64, (format.(test2[!,:Ninjas], precision = 0)))
predicted_test = parse.(Int64, (format.(ypredicted_test, precision = 0)))

2. Determining the size of your Vector using:

size(actual_test)
size(predicted_test)

This step is also a QA to make sure your
confusion matrix is mapped 1:1

3. Compute the Confusion Matrix

confusmat(56, actual_test, predicted_test)

Then as needed, you can apply Lighthouse
for items such as:

• Accuracy -- Lighthouse.accuracy(confusion:: AbstractMatrix)
• BinaryStatistics -- Lighthouse.binary_statistics(confusion::AbstractMatrix, class_index)

Hope this helps someone out there.

:koala:

QuBit (Jul 21 2021 at 00:27):

@Eric Hanson

Do you know how to apply
binary_statistics (C, precision)?
and reference it in the println()

I have added the function HERE

And added the binary_statistics function found right below the function
found at the link above.

Do I need to convert the matrix to a dataframe:

Bucket = DataFrame(binary_statistic(C))

Then reference the index as:

println("The Precision Score is", mean(Bucket[!, :precision]))

Yes, I have imputed the values. Might there be a
more elegant approach?

Thank you,

Last updated: Dec 28 2024 at 04:38 UTC