← Past problems

HiMCM 2018 · Problem set

The 2018 contest paired a data-heavy ranking problem (which roller coasters are objectively the best?) with a control-and-energy problem (how should a smart home decide when to heat/cool/ventilate?). Both have a strong feature-engineering flavour.

Contest datesNovember 7 – November 20, 2018 (14-day window)
Participation~730 teams worldwide [illustrative]
Problem ARoller Coaster — objective ranking from a feature dataset
Problem BCozy Smart House — smart climate-control schedule
Official results 2018 HiMCM results & commentary
Why 2018 is worth doing. Problem A comes with a real attached dataset (300+ roller coasters with steel/wood, height, drop, speed, length, inversions, year, etc.) — it's the closest HiMCM has come to a pure data-science task. Problem B is a great warm-up for any optimization-over-time problem, including 2021-A.

The two problems

Why this year is good practice

  • Problem A teaches feature engineering. Highest, fastest, longest aren't obviously commensurate — turning a heterogeneous table into a single score is the entire modelling exercise.
  • Problem B teaches dynamic optimization. Day/night, weekday/weekend, winter/summer — the schedule is naturally a small DP or LP.
  • Both have clean public data. RCDB.com and EPA / DOE residential energy data are tidy and free.