Try to do some basic machine learning on my uLogMe data, see what I can get?

[Naereen]

Try to do some basic machine learning on my uLogMe data, see what I can get?

[Naereen]

I need to first ask myself what questions I would like to answer with such statistical analysis.

[a1rb4Ck]

Hello, this so great to have your good working uLogMe fork! Don't have that much time to dig in, but here are some quickly thought ideas:

• Find your most productive pattern (hours, days, ideal work sequence, do you need breaks?, what's your best type of break? Away from keyboard or scrolling Social Networks?
• Find your less productive pattern (what interrupts you the most? Email? Kernel panic?)
• Predictive alert: trigger an alert if your about to procrastinate (Let's detect bad patterns: Professional mail answering -> perso answering -> social networks).
• Go deeper: add new inputs. Weather sensor. Location sensor. A live webcam input (front facing laptop cam) to infer in what posture, facial expression, hairstyle, you are the most productive...

Which ask a big question:

• What is a good productivity metric?
  On this, I like a bayesian productivity approach where you can assign percent scores to each app you use (ex: 0% for social networks, 100% for Terminal). This is what I'm used to with the macOs Timing app. Also App usage can be aggregate to Task, then you can assign score to Task.

Finally, we could all anonymize and upload our collected data, then concat to create the biggest worldwide dataset ever.. Let's answering crucial questions like, are the human more productive when it rains? Are french the most productive? So much coding to go.

[Naereen]

Hi,
Who there is a lot of ideas here… thanks!
Thanks for your interest in uLogMe!

I will definitely never have time to code anything as serious.
Plus, uLogMe is not so popular. Other time trackers have way more users, see for instance https://wakatime.com/

[Naereen]

I'll try to reply to your questions and ideas as soon as I have some free time…

[Naereen]

Hi @a1rb4Ck, sorry, of course I never had enough time to come back at this, and never thought about it again.

I'm just dropping here some links about a very impressive framework built by someone:
https://beepb00p.xyz/hpi.html, https://beepb00p.xyz/my-data.html, https://beepb00p.xyz/pkm-search.html
And by someone else: https://github.com/KrauseFx/FxLifeSheet