I am working with a data set that has both numerical/continuous data as well as categorical data. I know I can use one-hot encoding as a method when preprocessing the data, but ultimately I'd like to know what should be the ML algorithm to consider when wanting to predict a numerical/continuous value (the dependent variable), and all independent variables are a mixture of categorical and numerical data?
2 Answers
$\begingroup$
$\endgroup$
If I am not wrong you are trying to ask "are there any ml algorithms for regression task , that handles both numerical and categorical type characteristics while developing a model? "
Several ml algorithms like Decision tree regressor, Random forest regressor(which is an ensemble of several decision trees), SVM Regressor etc., The reason to consider one depends on how well they behave to foresay values.
$\begingroup$
$\endgroup$
1
As stated, the question is underspecified. You're essentially asking, "what modeling approach will always work?"
Use the XGBoost regressor and you won't go far wrong, even if you're unwilling to put much thought into model selection.
It will identify usefully informative features, whether continuous or (one-hot) categorical, and produce a plausible answer. Devoting more thought to the strengths of different model families and the peculiarities of your input data may suggest to you even better models for your (undisclosed) use case.
