In machine learning, regression problems deal with situations when the label information is continuous. This can be predicting the temperature for tomorrow, the stock price, the salary of a person or the rating of an item on an e-commerce website.
There are many models which can solve the regression problem:
- Ordinary Least Squares (OLS) is the usual linear regression
- Ridge regression and LASSO are the regularized variants of OLS
- Tree-based models such as RandomForest
- Neural networks
Approaching a regression problem is very similar to approaching a classification problem, and the general framework stays the same:
- First, you select an evaluation metric
- Then, you split the data into training and testing
- You train the model on training, tune parameters using cross-validation, and do the final verification using the held out testing set.