Which algorithm fits best for predicting continuous outcome from continuous and categorical inputs?



I have just started learning techniques to solve business problems and have gone through a table where they have suggested me to go with ANOVA if dependent is categorical and independent variables are continuous, Regression if both are continuous and Logistic if both are categorical.

Recently I have come across a problem (Kaggle - Bike Sharing Problem) where dependent variable is continuous but independent variables are mix of both categorical and continuous. Can you suggest which method should I use or do I need to learn new methods to deal with these problems.



I too have a similar question. My dependent variables are continuous and categorical. Output is Categorical. Is it ok to use both continuous and categorical variables as predictor variables?

Please advise



There is no specific way to determine that which Algorithms will work best. But there are few algorithms which you can try-

  1. Simple Multivariate Linear Regression - Just start with a simple linear regression to see how well it predicts. Do some feature creation and bi-variate analysis to understand the data well
  2. Regression Trees - This is a regression counter part of Random Forest. I have used this multiple time and it works brilliantly for me, but watch out for over fitting problem
  3. Ensemble Techniques - You can merge multiple techniques to get the best of each algorithm.

Hope this will help.