So much to tell you…
We had the amazing Amanda Scheirz – visit and regale us with tales of her data science exploits, as one of the few to go undercover in terms of gender into the world of Kaggle! See Amanda – “who for the longest time was top ranked female on Kaggle” (Anthony Goldbloom) in this FiveThirtyEight video. She generously shared her stories and lessons learnt with us, drawing insight from the different approaches i.e. hammer it with all the compute power you have or make sure you really understand your data and are asking the right questions. We all know that Kaggle is a world apart from the real one in some ways, but actually, listening to her talk, it seemed that if you remove the competition element and the finessing of accuracy to the nth degree, the distinctions blur…
What an inspiration, it was invaluable to be able to sit down and chat over pizza (thanks to H2O!) a glass of wine (thanks to The Data Analysis Bureau!) in the comfort and cosy intimacy of the Thoughtwork’s Soho kitchen (thanks guys!). But the biggest thank you simply has to go Amanda! The coolest lady in this realm that I’ve met, so giving of her time and expertise, and all in such a friendly manner – so welcome. But not just by me – by all the WIK ladies:
the most insightful and valuable talk by Amanda. A lot to take away from the evening! Class A meetup!”
“Amazing evening! Very insightful and inspiring talk by Amanda with very interesting examples.
“So welcome. Most useful talk I’ve been to/watched this year.”
“It was very useful and encouraging to see someone who had worked in this field for a long time and hear about her experience and advice.”
“This was the best meetup ever, thanks Amanda for coming to talk to us
Data science automation is not going away – but hang out for the creativity…we can leave the stuff that machines can do to the machines, but the stuff that the humans can do can that ever be automated, like the quantum leaps of imagination, innovation, originality? And if a computer can do them, can it explain its rationale or reasoning in a manner that complies with ethical regulations/restrictions? Big questions, and more to follow:
Women love him, guys wanna be him – what is it that Duncan’s got that made him……the first man allowed into Women in Kaggle? Big thanks must go too to Duncan for stepping in as our host these last two sessions! Perhaps we should pre-empt his ‘spy’ exposé with an exclusive interview….
Speaking of interviews, WIK got some fantastic coverage when I was interviewed earlier this month by Science Disrupt, check it out here:
This last Thursday, the Kaggle crew convened again to fashion some useful data products. We had an overview of Bokeh, with great contributions from Kim, but focused mainly on Shiny (the big word on the tips of tongues at EARL this year so I’ve heard on the grapevine). As we all know communication is one of the most vital elements of data science – if we don’t feed our insights back to stakeholders, it stagnates in the brain of the analyst, without fostering any proactive response, so we thought we’d explore some visualisation tools.
Building on the solid coding foundations of group members, like Suzanne, who supplied a ‘spend’ analysis by household – examining whose spend increases or decreases, with a view to ultimately identifying broader customer profiles in terms of spend behaviour – we transposed the analysis to Shiny, encouraging engagement with the raw data and empowering the user to pull out useful and timely information.
This has been a really good exercise and we’re planning on continuing the datadive into further sessions as it seems there’s a lot to be gained from working together as a group to explore a rich dataset from so many potential angles, using it as a framework for a variety of algorithms and tools.
Our focus on the regression problem tied in nicely with Kaggle’s House Prices competition that we looked at together at the outset, so we were able to extend our discussion to reviewing ways to up the leaderboard rankings of some members – talk about empowering!
The three hours fly by so quickly, sometimes we’ve just got stuck in and we’re saying goodbye, so I’m in the process of planning an all-day datadive in collaboration/partnership with Tableau, as part of their wider push in actively promoting diversity, so will keep you posted there…
Keep Kaggling till the next time,