Which Kaggle competition(s) are you working on?

I’m curious about which of the currently active Kaggle competitions everyone is working on. If you pick “Other”, it would be great if you could leave a note about which one you’re working on and why. Thanks!

  • Passenger Screening Algorithm Challenge
  • Statoil/C-CORE Iceberg Classifier Challenge
  • Cdiscount’s Image Classification Challenge
  • Spooky Author Identification
  • Dog Breed Identification
  • Other Active
  • Other Archived

0 voters


I feel that I need to really look back on the archived insurance ones and get those running simply enough that I can teach co-workers at AIG.


I’m also looking into the grocery forecasting one.

TIL the forums can do polls :slight_smile:


Looking at the Yelp competition.

Porto Seguro’s Safe Driver Prediction

Hi @KevinB, Was looking at this challenge too and i’m interested in taking part.

Am I correct in assuming that this challenge cannot be solved by CNNs? As the data will be product items with dates, sales and other item information and not images. If not CNNs can this be solved by some other deep learning algorithm?

Will be great if you can share more on this. New to this field - and my knowledge is limited to the first two lessons of the part1v2.


I would highly recommend going through the ml1 videos that Jeremy has been posting, but from my (limited) understanding of how to solve this type of problem, you won’t use a CNN, but it could still potentially be a Neural Network. My current plan is to join the files into one file and then use a random forest to see how that looks, but I have been more focused on the dog breed one up to this point.


Lesson 14 from course.fast.ai shows how to solve a nearly identical problem with neural nets.

1 Like

Guess we got lucky here :slight_smile: Really looking forward to next weeks lesson on working on structured data!


I should have included Corporación Favorita Grocery Sales Forecasting as an option in the poll :sweat_smile:

1 Like

Hey, I am also working on that one. Do you want to team up? Essentially I have been using the Rossman to get to:

md = ColumnarModelData.from_data_frame(path, val_idx, joined, yl, cat_flds=cat_vars, bs=128)

Then I fail because Im not understanding everything correctly.

Hi @mindtrinket . Yes, we can. Is today the last day for team mergers?

Can anyone translate the requirements for Rossmann’s submission file? I opened the sample file, it looks like this. I don’t think I have the right front for it.


That isn’t what that file should look like. Can you try redownloading that file?

Also Porto Seguro using the old Keras implementation. Works pretty well but not nearly as good as Boosting Algorithms and Ensembling.
Anyone up for Corporacion competition ?

1 Like

I’ve been working on that :slight_smile:

1 Like

Did anyone try and got good results on Porto Seguro Insurance competition using fastai ColumnarModelData?

I still haven’t been able to get the Rossman type code working correctly on it.

If you need someone to assist you, I would be delighted :smile: