Understanding model blueprints
DataRobot performs a lot of data transformations and hyperparameter tuning while building a model. It leverages a lot of best practices to build a specific type of model, and these best practices are codified in the form of blueprints. You can inspect these blueprints to gain insights into these best practices and also to better understand which steps were taken to build a model. To inspect the blueprint for a model, you can click on a model, go to the Describe tab, and then select the Blueprint tab, as illustrated in the following screenshot:
Figure 6.17 – Model blueprint
Here, you can see the workflow steps. As you can see, this blueprint is fairly simple. This is because gradient boost methods are very flexible and do not require a lot of preprocessing. Let's look at another model that did pretty well, the Generalized Additive2 Model (Gamma Loss) blueprint, as illustrated in the following screenshot:
