Dummy Regressors

Dummy Regressors

In Machine Learning, we start working on the problem, by developing a baseline model and then improving upon the model using other models or other Feature Selection or Engineering Techniques. There is a famous saying

---

All models are wrong, but some are useful.

---

Scikit Learn provides us with DummyRegressor, which involves using simple rules to make predictions and developing a baseline model for the regression problem at hand. It is not recommended for real time problems. The DummyRegressor class takes in a parameter str, which stands for strategy. This parameter decides which technique to use for making predictions. It can take one of the following values:

• "Mean": Always predicts the mean of the training targets.

• "Median": Always predicts the median of the training targets.

• "Quantile": always predicts a user-provided quantile of the training targets.

• "Constant": always predicts a constant value that is provided by the user.

In all these strategies, the model’s predict method completely ignores the input data.