This is a regression problem that involves predicting the price of e.g. aluminum, oil, strawberries. I have hourly and half hourly data for the weather and up to 10 different socioeconomic variables (all numeric), for roughly 8 years.

I want to make a machine learning model that extrapolates the price 7 days into the future. Would it be more suitable to use gradient boosted decision trees or neural networks? I'm leaning towards using trees because a shallow leaning method might be better for this problem.

  • $\begingroup$ This sounds like a traditional time-series forecasting problem. Between a neural network and a gradient boosted model I would recommend starting with a gradient boosted model. A neural network is more than capable for forecasting but for this problem, I would start with traditional regression models or forecasting models, test gradient boosting, and if you want then use a neural network. $\endgroup$ Jul 28 at 2:59
  • $\begingroup$ Thanks @David Gibson. I noticed in the sklearn docs that one of the disadvantages of decision trees is that, "Predictions of decision trees are neither smooth nor continuous, but piecewise constant approximations... Therefore, they are not good at extrapolation." (scikit-learn.org/stable/modules/tree.html#decision-trees) $\endgroup$ Jul 28 at 19:07
  • $\begingroup$ A tree-based model will not work on data with trends. $\endgroup$
    – 10xAI
    Jul 29 at 2:21
  • $\begingroup$ @10xAI Can you expand on your comment as an answer below? Thanks $\endgroup$ Jul 29 at 2:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.