Profile analysis for Boston and Cambridge

[shreyapandit]

This PR is a work in progress and will be mainly useful for providing a nice UI to look through the notebooks which pertain to the profile analysis.

[shreyapandit]

Looks like pylint has failed...will look into it

[bpben]

@shreyapandit there's a few pickle files you're merging here, which probably don't make sense to include. I think, generally, the only things that should be included here are the notebooks. And for that, I don't feel that we need much in the way of CI/coverge checks, it's experimental. Is that reasonable @j-t-t ?

[bpben]

@shreyapandit there's a few pickle files you're merging here, which probably don't make sense to include. I think, generally, the only things that should be included here are the notebooks. And for that, I don't feel that we need much in the way of CI/coverage checks, it's experimental. Is that reasonable @j-t-t ?

[j-t-t]

Certainly notebooks aren't included in the codecov metrics at this point. If @shreyapandit is writing a pkl file as part of the train_model, I might throw in a test for that but it would be simple/quick to do

[bpben]

Just took a look at the pickle, it looks like it's the same as the seg_w_predicted.csv that comes out of the model. Is that right, Shreya?

On Mon, Oct 22, 2018 at 10:03 AM Jenny Turner-Trauring < notifications@github.com> wrote:

Certainly notebooks aren't included in the codecov metrics at this point. If @shreyapandit https://github.com/shreyapandit is writing a pkl file as part of the train_model, I might throw in a test for that but it would be simple/quick to do

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Data4Democracy/crash-model/pull/207#issuecomment-431845008, or mute the thread https://github.com/notifications/unsubscribe-auth/ANMbgaOv_u78I8qUA68pu1lSHbW0YLg3ks5undBOgaJpZM4XdjwP .

[terryf82]

If we're aligned on focusing development on segment-level predictions in the future I think it would be good to update this notebook to reflect that. This still references a lot of week-level properties and I found it a little difficult to follow through for a city (Brisbane) that had segment-level preds.

[bpben]

Why don't we have a conversation about this this week. I think it probably doesn't make sense to try and expand development to other cities until we're comfortable with the methodology.

On Mon, Oct 22, 2018 at 6:08 PM Terry Franklin notifications@github.com wrote:

If we're aligned on focusing development on segment-level predictions in the future I think it would be good to update this notebook to reflect that. This still references a lot of week-level properties and I found it a little difficult to follow through for a city (Brisbane) that had segment-level preds.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Data4Democracy/crash-model/pull/207#issuecomment-432007949, or mute the thread https://github.com/notifications/unsubscribe-auth/ANMbgb_Mk5ieUXcnhtlty-98fSqTCg0gks5unkHKgaJpZM4XdjwP .

[codecov[bot]]

Codecov Report

Merging #207 into master will increase coverage by 0.03%. The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #207      +/-   ##
==========================================
+ Coverage   52.21%   52.24%   +0.03%     
==========================================
  Files          29       29              
  Lines        2959     2961       +2     
==========================================
+ Hits         1545     1547       +2     
  Misses       1414     1414

[shreyapandit]

Closing this for the moment, will reopen once we agree on a standard method to use for all cities.