New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Removing Bad Data #1432

Open
EricSchles opened this Issue Feb 24, 2017 · 8 comments

Comments

Projects
None yet
5 participants
@EricSchles
Contributor

EricSchles commented Feb 24, 2017

Currently CALC appears to have some bad data (hourly rates well over 1000 dollars an hour). So we want a strategy for removing such data.

Here is the strategy @toolness @jseppi have decided upon:

  1. a semantic classification of labor category as either likely high paying or likely lower paying

Example:

f("secretary") -> "low paying"
f("project manager") -> "high paying"

We'll be making use of an SVM (probably to do the text classification). We may also want to include some data from the prices we trust to reinforce this learning algorithm. Bringing in mixed data can be tricky. So we might need to do some preprocess encoding to the text in question.

As for the other data, I've made a general measure of spread and center. We can say if a price falls two spread lengths outside the center, that data is likely an outlier. Another anomaly detection technique we might consider is isolation forrests.

@jseppi

This comment has been minimized.

Show comment
Hide comment
@jseppi

jseppi Feb 24, 2017

Contributor

Anomaly detection seems like a great way to also detect if an uploaded price list (through Data Capture) has some weird pricing in it, and present it as a warning to the user!

"Whoa, your price list has a 'Project Manager' with an hourly rate of $1200. Is that correct?"

Contributor

jseppi commented Feb 24, 2017

Anomaly detection seems like a great way to also detect if an uploaded price list (through Data Capture) has some weird pricing in it, and present it as a warning to the user!

"Whoa, your price list has a 'Project Manager' with an hourly rate of $1200. Is that correct?"

@EricSchles

This comment has been minimized.

Show comment
Hide comment
@EricSchles

EricSchles Feb 24, 2017

Contributor

Yes, this looks good.

Contributor

EricSchles commented Feb 24, 2017

Yes, this looks good.

@toolness

This comment has been minimized.

Show comment
Hide comment
@toolness

toolness Apr 17, 2018

Contributor

Whoa, I just noticed this issue existed. FWIW, I did some exploration of anomaly detection recently on the calc-analysis repo.

Contributor

toolness commented Apr 17, 2018

Whoa, I just noticed this issue existed. FWIW, I did some exploration of anomaly detection recently on the calc-analysis repo.

@EricSchles

This comment has been minimized.

Show comment
Hide comment
@EricSchles

EricSchles Apr 17, 2018

Contributor

yea, its a real problem :/

Contributor

EricSchles commented Apr 17, 2018

yea, its a real problem :/

@jseppi

This comment has been minimized.

Show comment
Hide comment
@jseppi

jseppi Apr 17, 2018

Contributor

👋👋👋 @EricSchles

Contributor

jseppi commented Apr 17, 2018

👋👋👋 @EricSchles

@toolness

This comment has been minimized.

Show comment
Hide comment
@toolness

toolness Apr 17, 2018

Contributor
Contributor

toolness commented Apr 17, 2018

@hbillings

This comment has been minimized.

Show comment
Hide comment
@hbillings

hbillings Apr 17, 2018

Member

IT'S AN @EricSchles PARTY

Member

hbillings commented Apr 17, 2018

IT'S AN @EricSchles PARTY

@ericronne

This comment has been minimized.

Show comment
Hide comment
@ericronne

ericronne Apr 17, 2018

Contributor

wooo hey! 👋

Contributor

ericronne commented Apr 17, 2018

wooo hey! 👋

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment