Hi Maveryx, A solution to last week’s challenge can be found here. Welcome to the third and final lap of our Weekly Challenge based on the Inspire 2024 Grand Prix! As we reach the climax of this thrilling journey, we are excited to present a challenge on predictive analysis. This topic, not often featured in Weekly Challenges, promises to push the boundaries of creativity. Get ready to dive deep and showcase your predictive analytics expertise! Your task is to build a model to predict the top three podium finishers and compare the predicted versus the actual Silverstone podium finishers, and then identify the predicted racer who was not an actual podium finisher. Use only driver race averages for full races, not qualifiers in Japan, Qatar, and Qatar Sprint to train the model. Use the model you built to score any drivers with full race data from Silverstone and determine the three most likely podium finishers. Then, identify any of those three drivers who did not actually make the podium at Silverstone. Minimum lap counts per race to determine full race data: * Japan: 53 * Qatar: 57 * Qatar Sprint: 19 * Silverstone: 52 The tasks to accomplish your objective include: * Create a training dataset using the Race Data Japan_Qat_Quali_Sprint dataset for your model. Be sure to exclude qualifying races and drivers who did not complete a full race using the minimum lap counts above. * Using this training dataset, build a Logistic Regression to estimate the likelihood a driver will finish in the top three on the podium.* Use any Avg variables as possible predictors and nothing else. * Use a Stepwise tool with the default configuration to determine the best predictor variables. The Stepwise tool will automatically determine the final model variables for the model. * Score any Silverstone drivers with full race data using the Stepwise output and identify the top three most likely podium finishers. * Compare the predicted podium finishers versus the actual Silverstone podium finishers. Find the predicted driver who did not finish on the actual podium. Feel free to use the hints provided within the workflow. Need a refresher? Review the following lessons in Academy to gear up. * Predictive Modeling * Predictive Analytics Fundamentals * Creating a Predictive Model Good luck! Challenge_430_start_file.yxzp Challenge_430_solution_file.yxzp

Forgot to upload. Spoiler Challenge_430_start_file-Qiu.yxzp

I think my favorite thing about this challenge was discovering that actually nobody is all that likely to win a Formula 1 race??? challenge_430_solution.yxzp

nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

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

AYXAcademy

Hi Maveryx,

A solution to last week’s challenge can be found here.

Welcome to the third and final lap of our Weekly Challenge based on the Inspire 2024 Grand Prix! As we reach the climax of this thrilling journey, we are excited to present a challenge on predictive analysis. This topic, not often featured in Weekly Challenges, promises to push the boundaries of creativity. Get ready to dive deep and showcase your predictive analytics expertise!

Your task is to build a model to predict the top three podium finishers and compare the predicted versus the actual Silverstone podium finishers, and then identify the predicted racer who was not an actual podium finisher.

Use only driver race averages for full races, not qualifiers in Japan, Qatar, and Qatar Sprint to train the model. Use the model you built to score any drivers with full race data from Silverstone and determine the three most likely podium finishers. Then, identify any of those three drivers who did not actually make the podium at Silverstone.

Minimum lap counts per race to determine full race data:

Japan: 53
Qatar: 57
Qatar Sprint: 19
Silverstone: 52

The tasks to accomplish your objective include:

Create a training dataset using the Race Data Japan_Qat_Quali_Sprint dataset for your model. Be sure to exclude qualifying races and drivers who did not complete a full race using the minimum lap counts above.
Using this training dataset, build a Logistic Regression to estimate the likelihood a driver will finish in the top three on the podium.
1. Use any Avg variables as possible predictors and nothing else.
2. Use a Stepwise tool with the default configuration to determine the best predictor variables. The Stepwise tool will automatically determine the final model variables for the model.
Score any Silverstone drivers with full race data using the Stepwise output and identify the top three most likely podium finishers.
Compare the predicted podium finishers versus the actual Silverstone podium finishers. Find the predicted driver who did not finish on the actual podium.

Feel free to use the hints provided within the workflow.

Need a refresher? Review the following lessons in Academy to gear up.

Good luck!

AYX-Maveryx_Micro Identity-Maveryx Academy_wordmark.png

Challenge_430_start_file.yxzp

Challenge_430_solution_file.yxzp

Predictive Analysis

Expert

Difficult

Accepted answers

All comments

ahsanaali

Spoiler

If you skip PitStopAvg predictor variable, then predictions are 100% accurate. How do I know that....well that's what I missed initially and spent good 10 mins finding why prediction were 100% accurate :-).

Challenge_430_Solution_AAA.yxmd

ggruccio

Solved it! Sorted on the score for "Yes".

Spoiler

Challenge_430_gg_finish.yxzp

mmontgomery

C430

Spoiler

I made lots of small mistakes, but I eventually got it in 9 minutes

c430.yxzp

AkimasaKajitani

My solution.

Spoiler

Today, I misunderstood very much. So it took too much time. In other words, the predicted results show very little difference between all the racers.
スクリーンショット 2024-06-25 001054.png

Challenge_430_start_file_AK.yxzp

aatalai

More predictive challenges please

aatalaiChallenge_430_soltuin.yxzp

RolandSchubert

Spoiler

Challenge_430_solution_rsc.yxzp

ed_hayter

Spoiler

good refresher on predictive

Challenge_430_start_file.yxzp

Qiu

This will take me a while. 😁

estherb47

Fun predictive exercise!

Challenge_430_EHB_completed.yxzp

Chris-Nienart

A good toy model for learning to configure predictive tools.

With so few observations, there really isn't much predictive power and very high likelihood of overfitting.

In my solution, I also added some model comparison tools to the analysis.

Spoiler

Model Comparison

Spoiler

By accuracy alone, both models appear to be doing a good job, however, they are both quite poor predictors.
Both models predict 'No' for every observation. The Recall is 0% and Precision is NaN (0/0).
Hence Accuracy_Yes is 0 and F1 (a function of Precision) can't be computed.

My-solution-430.yxmd

Luke_C

Spoiler

Challenge_430_start_file.yxmd

sfarnham1

Spoiler

Challenge_430_start_file.yxmd

RashedDS

Spoiler

Challenge_430_RashedDS.yxzp

olga_strubbe

A great challenge - this was a good illustration of using predictive tools. Thank you.

Challenge 430_OS.yxmd

tammybrown_tds

Spoiler

Challenge_430_tbrown.yxmd

alexnajm

Finally done! Definitely not beating Molly's time 😂

Spoiler

Challenge_430_start_file_Alex Abi-Najm.yxzp

DaisukeTsuchiya

Spoiler

Challenge_430_start_file_DT.yxzp

LindonB

Good refresher on how to configure a couple of these tools!

Spoiler

Challenge_430_Solution.yxmd

RWvanLeeuwen

Interesting predictions...

Spoiler

So we use the post race statistics to predict? Smells like post-hoc analysis in here

with scores that low, you'd expect that everybody loses in this race

Challenge_430_start_file - RWvL.yxmd

alacoume

A solution:

Spoiler

Challenge_430_start_file.yxmd

AGilbert

A Ferrari P4 at Silverstone? We will see in a couple of weeks!

As always with the predictive tools, looking forward to seeing the solution next week and understand data science more.

Spoiler

Challenge_430_ag.yxzp

balajilolla2

Spoiler

Solution Attached

Screenshot 2024-06-26 164311.jpg

Challenge_430_start_file.yxmd

jeprime

Spoiler

Challenge_430_jepsolution_file.yxmd

LBishop011

Fun one! Hadn't used the Predictive tools before so good to learn.

Challenge_430_start_file.yxmd

Kenda

Spoiler

Challenge_430_start_file.yxmd

mkeiffer

My solution. This was a fun one!

Spoiler

Screenshot 2024-07-03 at 10.50.28 AM.png

Challenge_430_mpk.yxmd

JoachimCaronTIL

Here is my solution

Spoiler

Challenge_430_JC.yxmd

KevinHarrison

I was a little confused with the instructions because it said not to use the Sprint races but then said to use them. Anyhow, I got to the solution it seems.