0
$\begingroup$

Does Arima support the usage of categorical variable? Some ways to get it working can be using one-hot encoding to represent categorical variables, but I am not sure how good it is.

$\endgroup$
  • $\begingroup$ ARIMA models are univariate models (one time series only). You cannot fit categorical features as a result because how are you supposed to define categories across a single time series? ARIMA models can handle time related exogenous variables (promotion/no promotion, recession/no recession), which are often indicated using indicator variables. However, problems may arise during forecasting when you do not know the future values of some of your predictors (forcing you to forecast predictors as well, making error potentially worse) but in general it can still lead to improvements. $\endgroup$ – aranglol Aug 22 '19 at 19:20
  • $\begingroup$ If you want to incorporate categorical features, you need to consider fitting a single model to multiple time series that differ by some category. In this case, ML methods start to become potentially more attractive then traditional statistical ones, at least that is what performed the best in recent competitions that allowed external regressors. $\endgroup$ – aranglol Aug 22 '19 at 19:27
0
$\begingroup$

One-hot encoding is the way. For ARIMA models, no other configuration is possible. Whether it is good or not, it depends on the quality of your data, your variable choice, and a correct model specification. Unfortuntately, we can't say a priori whether the result is going to be good or bad.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.