Challenge #430: Inspire 2024 – Grand Prix (Lap 3)

[patrick_digan]

Spoiler

[ARussell34]

I found the driver!

[Reesetrain2]

My submission!

Spoiler

 

[Garrett_Stoker]

Regex just to make it interesting.

Spoiler

[Erin]

Spoiler

[DawnDuong]

Good refresher of the Stepwise and LR tools. Thank you for sharing this challenge.

[Caramel8]

Spoiler

I learned that the join tool will somehow create an error with the header matching in the Score tool. 

[Alfie_King1]

Spoiler

 

[OllieClarke]

Here's my solution which matches the output, and some thoughts on oversampling

Spoiler

I thought that If the Podium finish column is only there in ~24% of records, should we not be oversampling here?
23.68% Yes

I tried it and it broke everything, (I think because there were too few records left over from the undersampling)
You get a 100% accurate logistic regression (which warns you about the lack of rows), but after scoring no drivers are predicted to podium.
No one gets a podium

If we do oversample though (rather than using the tool)
Oversampling the "Yes" rather than undersampling the "No"

We get a more accurate logistic regression than the basic workflow (although less accurate than the oversample one)
oversampled logistic regression
We also get a model that outputs the actual 3 podium finishers as the 3 most likely to podium (with Leclerc 4th most likely)


The oversampling section might be too much for a grand prix leg, and there's not a lot of data anyway, but is oversampling the correct approach here?

[Bobbyt23]

Good practise with predictive tools. Couldn't do it under pressure on stage though!!