After preparing and exploring the data, the next step will be to build the model. The dataset has been divided into train and test in a previous step. It is crucial to define a clear methodology to select an algorithm, tune it and train the model. To keep it simple, let’s define a simple but effective methodology.
- Define a metric or metrics for the evaluation.
- Select a set of algorithms to be evaluated.
- Tune the hyperparameters of all algorithms using cross-validation and the training dataset.
- Use best hyperparameters of all algorithms to train a model using the whole training dataset.
- Final evaluation.