IBM SkillsBuild Hydropower Climate Optimisation Challenge
Helping the World
$3 000 USD
Completed (~1 year ago)
Prediction
Forecast
1233 joined
462 active
Start
Mar 03, 25
Close
Apr 13, 25
Reveal
Apr 14, 25
Boost Your Score! Starter Notebook with 7.17 RMSE Public LB
Notebooks · 3 Apr 2025, 21:35 · 8
Hey everyone! 👋
I noticed that many of you were looking for more insights after the webinar, so I decided to share a starter notebook that improves performance. This one achieves a 7.17 RMSE, which is a step up from the baseline. 🚀
🔗 https://www.kaggle.com/code/dukekojokongo/ibm-skillsbuild-improved-starter-notebook
What's Inside?
✅ Feature engineering tricks to improve model performance
✅ A better handling of missing values and zero values from months 1 to 6
✅ Optimized model parameters for better forecasting
I hope this helps! Feel free to test it, tweak it, and share your thoughts. Let’s push for even better scores together! 💪🔥
Credits to @MuhammadQasimShabbeer for the uploaded dataset
Don't forgot to upvote the notebook :)
Nic and neat notebook @CodeJoe , i appreciate your effort and help
Thank you for the kind words @Knowledge_Seeker101
Hi @CodeJoe! I'm still struggling with the best strategies on validation and feature engineering and your notebooks are helping a lot! I've tried to reproduce your example but the "complete_data" dataframe in your notebook has 32906 rows while when I try to reproduce it with the raw data from the "Data" page, it turns out to have 136409 rows, almost 4 times more rows. Have you done any pre-processing? E.G.: filtering some periods of dates or some combinations of consumer_device/user_id?
@silvaemqap Yes I basically dropped the zero values to get a better score with gradient boosting models. All of it is explained in the notebook.
Good job @CodeJoe - also trying to make peace between the massive gaps between cv scores vs p.leaderboard in my experiments.. Kind of competition that may give us another earthquake just before I recovered fully from the cassava root estimation one..
You can try a sliding window validation @silvaemqap and see how will turn out if not yet....
I understand your plight😂. we hope for the best.
Thank you very much @CodeJoe, you just made my day, I am happy. Your notebook is very neat and very comprehensible. You are a great data scientist, keep it up.
@RareGem Thank you for the kind words. I really appreciate it.