I got a great opportunity to attend pydata conference ( thanks to pyladies for supplying the free ticket and my manager to let me attend it :))
It was a cool conference with great workshops, talks and presentations! (liked it more than the pytexas conference last year :)). Shout out to the organisers!
Heard many great/innovative talks about data analysis and predictions and machine learning based on that analysis-prediction model and the privacy concerns that come with open data at Pydata-dallas conference .. but not much on how to tackle/prevent the false predictions, their negative impacts, how are we doing the reality check on these "click-based analysis"
How do we predict that out of these 20 people only 2 are gonna show up on your meetup .. based on theit history ?
How do we predict and take out clicktivists from activists .. to make sure we are not grossly overestimating out support in this cause .. to be able to plan accuratly ?
How do you make sure doing a search on pressure cooker .. will not land you in some "special files" because of some social media analysis algorithm's incorrect prediction ?
How do we make sure that X's brand/popularity is not just "click popularity" ?
How do we establish actual learning assessments with Moocs?
Now the next question that comes to my mind is ..
Is it really that easy to manipulate the market sentiments by just some clicks?
Many invest in shares/funds e.t.c largely based on the market sentiment , brand's perception .. is it really that volatile ?
No comments:
Post a Comment