Julia

At a Nextmv tech talk a couple weeks ago, I showed a least absolute deviations (LAD) regression model using OR-Tools. This isn’t new – I pulled the formulation from Rob Vanderbei’s “Local Warming” paper, and I’ve shown similar models at conference talks in the past using other modeling APIs and solvers. There are a couple reasons I keep coming back to this problem. One is that it’s a great example of how to build a machine learning model using an optimization solver. Unless you have an optimization background, it’s probably not obvious you can do this. Building a regression or classification model with a solver directly is a great way to understand the model better. And you can customize it in interesting ways, like adding epsilon insensitivity. ...

Note: This post was originally written using Julia v0.2, GLPK, and Hedonometer data through 2014. It has been updated to use Julia v1.11, HiGHS, and data through May 26, 2025. Hedonometer popped onto my radar a couple weeks ago. It’s a nifty project, attempting to convert samples of words found in the Twitter Gardenhose feed into a time series of happiness. While I’m not a computational social scientist, I must say the data does have a nice intuitive quality to it. There are obvious trends in happiness associated with major holidays, days of the week, and seasons. It seems like the sort of data that could be decomposed into trends based on those various components. The Hedonometer group has, of course, done extensive analyses of their own data which you can find on their papers page. ...