The second workshop picked up, where the first ended. A quick recap from three weeks before: Dóri showed us how to handle a CSV dataset in pandas, how to sort, and count values, rename, remove columns, and tricks like this. She used a Pokemon dataset, which wasn’t just funny, but easy to follow.
So after we get familiar with pandas, it’s time to look into the Machine Learning part. I (Balázs) was the one who prepared for this event with an Unsupervised Learning problem.
Our venue was the oktatoterem.com again, and our room was designed for 18 people. It was so great to see, that more than 20 of you came to spend this evening learning Python with us! We were packed, but we managed this with extra chairs.
I made a Jupyter notebook with DataCamp’s “Musical Recommender” sample dataset. I left out some parts of the notebook, to think together, and I got some interesting questions, and ideas about where to go next, or how to evaluate this task. Our goal was to recommend artists to users with the same taste.
For example, if you listen to a lot of The Beatles, it recommends you Beach Boys, or Bob Dylan. To achieve this we reshaped our data, put it in a csr_matrix, scaled it, and reduced the dimension from 111×500 to 111×20. After these steps we could try various artists and discuss, if our model is good enough.
Thanks for everyone to show up!
As always, the notebooks (the full, and the one with missing parts) are on our GitHub:
You can join us on our meetup page: