Settling in for the last talk of the day at #YOWData: is the 370 the worst bus in Sydney?
Graphs are everywhere! We can’t use ML techniques that assume independence since the entities are interconnected. Andrew Docherty at #YOWData
Andrea Burbank retweeted
I was able to grab only part of the talk. Loads of great insights by @lindamciver #YOWData Bring your datasets, sponsorship and teachers to @DataSciAu
Amazing arguments by @lindamciver: “We need our kids to be data literate so that antivax isn’t a thing and so that we can make progress.” Building resources to help teachers learn to teach data science and computation at ADSEI. #YOWData
Excellent talk on human irrationality and data: “It turns out you can stare yourself into honesty, which I think has interesting implications for the selfie culture.“ #YOWData @lindamciver
The hazards of s3 according to @rohandhupelia: “You can just keep throwing data in there and not realize how much it’s building up!” Indeed. #YOWData
Lessons learned at Atlassian per @rohandhupelia: success came from realigning the data engineering team to align with customer needs.
Managing a self-serve analytics architecture at Atlassian: impose data architecture via naming conventions, and use an event bus to manage incoming connections. @rohandhupelia #YOWData
On data governance from @FCTweedie #YOWData: “I seem to pop up at a lot of meetings and say, ‘You seem to be collecting some data. Would you like some governance with that?’” How can we build data ethics and governance earlier into the product process?
“If you’re managing toward a known outcome, you’re not doing data science. In data science, you’re always dealing with unknown unknowns.” @cargomoose #YOWData
The business only needs data science if it is facing real competition, threats, uncertainty, and change. KPIs for data science outcomes exist and are sensible (accuracy, not models per week). The CEO has questions that data can answer. @cargomoose #YOWData
On data visualization: “My whole goal was to lure you into an explanation of why pie charts are crap.” -Hercules Konstantopoulos #YOWData
On human disagreement in sentiment analysis: “If language were precise, there would be no art, no poetry. No one wants to say exactly what they mean.” #YOWData @nomiddlename
“Sentiment analysis is extra tricky because customers stubbornly refuse to use the complete marketing term when referring to a given product.” @nomiddlename #YOWData
Fascinating talk from Dana Bradford on using sensors of motion, humidity etc to create a picture of seniors’ activity: showering, dressing, and more. Only hazard: subjects breaking sensors to get more visits from handsome technicians! #YOWData
We can easily see the cessation of sensor activity after Mrs Elle’s stroke, but to preserve her quality of life we need to identify stroke warning factors sooner. Sensors show reduced activity, low body temp, no meal prep: could we save her life with sensors? #YOWData
