CH Robinson: Trucking Costs
As part of the IMA Math-to-Industry Boot Camp VI during the summer of 2021, I spent three weeks in a group of six tackling a problem posed by C.H. Robison, a third party logistics company based in Minneapolis, MN. They challenged us to find a good method for predicting the price of trucking both refrigerated (reefers) and non-refrigerated (vans) between North American markets over the span of long-term contracts lasting anywhere from 3-12 months.
We made use of univariate models: SNaive, STL, SARIMA, TBATS, and Prophet; as well as multivariate models: SARIMA, VAR, and Prophet. We also developed a new set of strategies to blend expert knowledge about an industry into the prediction models, including via case studies of various markets, exogenous variables for market codependency/covariance, and tipping point analysis.
GitHub Repository
I have included much of my code (mostly written in R) in my personal GitHub repository:
