Feature Engineering is the process of identifying features in a given dataset, creating new features based on insights from data and preparing the features relevant to the models selected. The techniques used in this process are called transformers.
This is usually an iterative, time-consuming process for data scientists and often takes the majority of their time when building machine learning pipelines.
Automatic Feature Engineering
H2O Driverless AI automates the entire feature engineering process:
- Detect relevant features in a given dataset
- Find the interactions within those features
- Handling missing values
- Derive new features from data
- Compare the existing and the newly generated features
- Show the relative importance of each of these features
Outcome of H2O Driverless AI automatic feature engineering:
The input features are now transformed into meaningful values that the machine learning algorithms can easily consume.
Automatic Feature Engineering Using Recipes
Driverless AI users can really take advantage of the flexibility it offers in the feature engineering process via built-in transformer recipes, an open catalog of recipes and using the BYO functionality.
- Built-in Recipes
- Open catalog of recipes
- Bring Your Own Recipe
Driverless AI has X built-in feature engineering transformers.
Driverless AI has a growing list of over 60 feature engineering transformers in the open-source recipe catalog.
Driverless AI is an open and extensible Machine Learning optimization platform. This extensible is delivered by the Bring Your Own Recipe (BYOR) architecture.