linking-writing-processes-to-writing-quality

[golden moat]

Just started this today pretty fun so far

[golden moat]

So much more to work on, but a nice start for the contest

[topaz whale]

Hi All, Can someone explain up_event and down_event. Having a hard time wraping my head around it.

[golden moat]

@topaz whale when someone enters something on keyboard I believe

[topaz whale]

down_event - The name of the event when the key/mouse is pressed
up_event - The name of the event when the key/mouse is released

So have they like given just the keys being pressed and released ?

[golden moat]

correct

I found you can use these for multiple calcs

I dont have the best model yet, but its working

[lament dove]

Hi guys, I just joined the competition and I would like to know if there is a particular format for the notebooks we must submit. Can we get something like a sample_submission_notebook.ipynb showing what should absolutely be in our submission notebook in order to test our solution (if ever there are criteria for this).

[golden moat]

Still looking for a team if anyone is interested

[lament dove]

People are kind of quiet on here 😂

[golden moat]

Doesn’t end for 3 months so still a long time

[lament dove]

Hey everyone, I’m a bit confused about the real nature of scores in our data. I firstly thought that there would be 7 classes (int 0 to 6) but it seems that there is currently 12 classes in the training data (0.5 to 6 always increasing by 0.5). Is it possible that there will be other values in the data that will be used to test our solutions like 0 or 1.2 etc. I’m wondering if we should think of scores like continuous values cause it was written that their test data will have some features we don’t have in our training set.

[viscid ravine]

I think there would be only 12 classes in the test data. You don't need to worry about this if you take it as a regression task.

[lament dove]

Sounds great!👍 thanks!

[lament dove]

Hey! I’m losing about 30 thousands of rows each time I read the train_logs.csv file with pandas.read_csv(). Am I the only one who is encountering this?

[vast girder]

I usually use Polars
You can also try this

[lament dove]

Thanks🙏🏿

[topaz skiff]

https://www.kaggle.com/competitions/linking-writing-processes-to-writing-quality/discussion/444592#2487073

?

[inner depot]

so why do they include down events for characters like ''t", "s"? Doesnt that defeat the purpose of the hiding the characters with q?

[topaz skiff]

Yep, exactly, Need clarity on what these characters would denote. @organizers

[languid plover]

this dataset is not working properly on google colab, I am trying to check the shape of data but always I am getting different values

[languid plover]

I am gaining data in every rerun , still don't know what is happening, also not working properly in kaggle notebook😆

train_log.shape == (8405898, 11), is this correct?

[lament dove]

I got a little bit less rows than you with my local machine but there is even more entries in the csv file

[topaz skiff]

I can see same shape on my Kaggle Notebook

[languid plover]

since it is a big data , try to reduce size or use vaex , then it will work properly

[languid plover]

yesterday , code was working fine but today i am getting error ,AttributeError: module 'vaex' has no attribute 'ml'

help

[haughty crest]

Hi Ryan,
I can join your team ! I have just started exploring the challenge and still a long way to go! And this would be my first Kaggle competition

[queen gyro]

hey y'all, i am currently taking a class in data science (i have a little bit of coding knowledge, can read it and understand the logic behind it but can't code completely on my own yet (but thats where gpt, etc. come in)). as our assignment we have to create a colab notebook to solve problems like predicting grades based on regression, detecting fake news using nlp and classification and so on. my professor told me this project here would be fitting as well. would you say thats a doable thing or is it way too complex for practically a beginner? of course i dont have to rank high in the competition but a fairly good notebook is expected. whats y'alls thoughts?

[unborn quarry]

Hi @golden moat would you like to form a team? I did just started with the competition
currenlty I'm on 558th position ( it's not great since I've used just single pycaret pipeine as of now )

let me know if you are up
my profile: https://www.kaggle.com/somesh88

[unborn quarry]

I think you can take a look at public notebooks ( you can get idea about the competitions from those )

[polar crypt]

This comp has been so quiet LOL

[dreamy gale]

More feature I engineer, the wrose result is

Any available suggestion?

[lament dove]

Is there anyone here looking to merge teams?

[calm bear]

If you add new extracted features and get worst results (CV) then some of your new features are harmful , so at this step you can apply some feature selection methods https://www.kaggle.com/code/prashant111/comprehensive-guide-on-feature-selection. Or advanced methods e.g. permutation importance , LOFO, Boruta, Null Importance , also you can use SHAP to interpret your results in combination with feature selection.

[north sonnet]

wow, it's really a good notebook for me, thank you for your sharing

[formal apex]

just a question, why is this alert still up there?

[tall sapphire]

@formal apex Thanks for the report, looks like this is a bug. We're going to fix it now.