Feedback - Online Hackathon - Predict the gem of Auxesia for Magazino!


#1

So you all saw the fun in finding the hidden gems in data. We hope you had an awesome experience doing so. We think that it was a fabulous problem and throbbing competition for the top spot - may the best (and the most lucky) data scientist win.

Now, coming to the point, we need some feedback on how was the experience? What worked and what would you like to see more of? What would you want to change? Any thoughts on format? marketing? communication? Do let us know.

It will improve us become better in the next version. Looking forward to hear from you all

Kunal


#3

First of all… The entire experience of the hackathon was awesome :smiley: I liked mostly the way it was executed.

In terms of improvement we can upgrade the design and performance of the LB page/web-app/website.
If you wish you can start a open-source project on Github. We would love to contribute. :smiley:


#4

Hi @kunal, First of all thanks for organizing this hackathon. The overall experience of the hackathon was good. If You could request top hackers to share their approaches and the reason behind choosing the particular approach, then it would be good for all.

Thanks in advance


#5

The experience was great. Got to learn so many new things. :smile:

As Vikash already mentioned, the looks and performance of Submission page can be improved but its not that serious.

You can also start a thread where the participants can explain and discuss their solutions, after the competition is over!


#6

@gauravkantgoel

Check this out:

http://discuss.analyticsvidhya.com/t/revealing-your-approach-predict-the-gem-of-auxesia-for-magazino/2278


#7

@kunal I had a suggestion/observation for your consideration. Going forward can we also have contests which are not so short term. 48 hours is a bit of a stretch for non experts like me. Beginners will struggle to even get a decent submission in. I believe if a contest runs for lets say a week, then it would have the following benefits

  1. More participation and engagement in the forum. I felt the forum conversations were a bit limited. But even with that, I learnt a few new things!
  2. Better grasp and absorption of the business context at hand. Short time frames cause us to not care about the meanings of terms like negative polarity, what would it mean if keyword min, max, mean are same, why is the shares so skewed etc etc. Instead we tend to treat the numbers as inanimate numbers to be fed into complex models.

#8

Great! Thank You… :slight_smile:


#9

I too think a github project would be a great idea. :smile:
I am also new into data science and still learning. Like @manizoya_1 said, 48 hours is a bit difficult on beginners. It would take us longer time to come up with a decent solution.


#10

Excellent Hackathon though i could not come up with submission due to being a beginner. I think it would be a good idea if the top 10 leader board could share the code in Github.


#11

#12

Was the final ranking based on 100% of the test set ?

I could be wrong, but I think it should have been only on the totally unseen 60% of the test set.