​
The Anatomy of a Feature

A feature consists of a value and an associated entity value. These features are defined based on registered data sets within Featureform. The entity effectively serves as a primary key or index for the feature values. While some features statically describe an entity (e.g., a restaurant’s zip code), others change over time. When features exhibit temporal variations, it becomes crucial to employ point-in-time correct feature values in training sets to prevent data leakage and enhance model performance.

​
Features without a Timestamp

Certain features remain relatively static. Examples include the category to which a product belongs. In such cases, the value of the feature is indexed by the entity, allowing you to look up the feature by entity.

To define a feature of this type, you add a Feature to an entity. The feature’s first parameter specifies the Featureform dataset, indicating the entity column followed by the value column in the form dataset[[entity_col, value_col]]. Optionally, you can set the variant. The type can take on one of the following values: ff.Int, ff.Int32, ff.Int64, ff.Float32, ff.Float64, ff.Timestamp, ff.String. Since features are typically served for inference to your trained model, it’s essential to specify the inference store for materializing the feature.

Example:

import featureform as ff

@ff.entity
class User:
  age = ff.Feature(dataset[["user", "age"]], variant="simple", type=ff.Int, inference_store=redis)

​
Features with a Timestamp

Some features exhibit changes over time, such as the highest priced item a user has purchased. In such cases, the feature’s value is indexed by the entity, and you can access the feature’s value as it existed at a specific timestamp. This is pivotal for creating point-in-time correct training sets.

To define a feature of this type, you add a Feature to an entity. The feature’s first parameter specifies the Featureform dataset, encompassing the entity column, value column, and timestamp column in the form dataset[[entity_col, value_col, timestamp_col]]. Optionally, you can set the variant. The type can take on one of the following values: ff.Int, ff.Int32, ff.Int64, ff.Float32, ff.Float64, ff.Timestamp, ff.String, ff.Bool. Since features are typically served for inference to your trained model, it’s essential to specify the inference store for materializing the feature. To maintain point-in-time correctness, only the most recent entity-feature pair is retained in the inference store.

Example:

import featureform as ff

@ff.entity
class User:
  age = ff.Feature(dataset[["user", "top_item", "timestamp"]], variant="simple", type=ff.Int, inference_store=redis)

​
Serving Features for Inference

Once a feature has been defined and applied, it will be materialized into the inference store for serving. The Featureform client provides a features method to serve your features.

Example:

client.features([("age", "simple")], entities={"user": id})

This retrieves the most recent value of the feature for the specified entity.

​
Building Training Sets with Features

A training set consists of a label joined with a set of features. You can define it as follows:

ff.register_training_set(name, variant, label=(name, variant), features=[(name, variant)])

When both the label and feature have a timestamp, Featureform automatically generates point-in-time correct training sets for you. You can then serve it as a dataframe or via a streaming method:

client.training_set(name, variant).dataframe()
EntitiesEmbeddings