This is a guide to
Triage, a data science workflow tool initially developed at the Center for Data Science and Public
Policy (DSaPP) at the University of
Chicago and now being maintained at Carnegie Mellon University.
Triage helps build models for two common applied
(a) Early warning systems (EWS or EIS), (b) resource
prioritization (a.k.a "an inspections problem") . These problems are
difficult to model because their conceptualization and and
implementation are prone to error, given their multi-dimensional,
multi-entity, time-series structure.
This tutorial is in sync with the latest version of
triage. At this moment v4.1.0.
How you can help to improve this tutorial
If you want to contribute, please follow the suggestions in the triage’s github repository.
What is in the name?#
There is a famous (and delicious) chinese duck restaurant in Chicago, we love that place, and as every restaurant in Chicago area, it gets inspected, so the naming is an homage to them.
Who is this tutorial for?#
We created this tutorial with two roles in mind:
A data scientist/ML practitioner who wants to focus in the problem at his/her hands, not in the nitty-gritty detail about how to configure and setup a Machine learning pipeline, Model governance, Model selection, etc.
A policy maker with a little of technical background that wants to learn how to pose his/her policy problem as a Machine Learning problem.
How to use this tutorial#
First, clone this repository on your laptop
git clone https://github.com/dssg/triage
Second, in the cloned repository's top-level directory run
This will take several minutes the first time you do it.
After this, you may decide to do the quickstart tutorial.
Before you start#
What you need for this tutorial#
Note that if you are using
GNU/Linux you should add your user to the
docker group following the instructions at this
At the moment only operative systems with *nix-type command lines are
supported, such as
MacOS. Recent versions of
Windows may also work.