featuretools
Automated Python feature engineering library using Deep Feature Synthesis on relational datasets.
Transform multi-table datasets into machine learning feature matrices using Deep Feature Synthesis.
Featuretools is a Python library that automates feature creation from relational and time-series datasets. It uses Deep Feature Synthesis to aggregate and transform relational data into a flat feature matrix ready for machine learning models. Data scientists looking to replace manual feature creation in Python pipelines should use it.
What it is
The library maps relationships across multiple tables using EntitySets and applies transformation and aggregation primitives. It automatically generates feature vectors for a target dataframe by combining nested calculations over relational structures. You can also define custom primitives or extend functionality with optional NLP and Dask extensions.
Who it's for
This tool suits Python data scientists and machine learning engineers working with multi-table relational databases and transactional data. It is not designed for non-programmers who require a graphical user interface or tabular point-and-click tools. It also requires raw data to be structured into relational formats before generating features.
Key features
Deep Feature Synthesis Engine
Automatically builds nested aggregation and transformation features across related tables.
Custom Primitive Definitions
Allows you to write bespoke Python functions when built-in transformation rules are insufficient.
NLP and Scaled Extensions
Supports natural language processing primitives and Dask integration for parallel execution.
Up and running in 3 steps
Our honest verdict
What's great
- Automates relational feature creation directly from multi-table datasets.
- Offers customizable feature primitives for specialized domain calculations.
- Integrates with Dask to execute feature generation across parallel compute clusters.
- Includes built-in demo datasets and tutorial notebooks for fast evaluation.
Where it falls short
- Generating deep features across large datasets requires significant memory and CPU resources.
- Requires upfront work to correctly define dataset schemas and table relationships in Python.
- May generate hundreds of low-value features that demand subsequent feature selection.
Pricing & license
Featuretools is free to use under the BSD-3-Clause license. The primary costs come from the compute hardware and memory needed to run feature extraction on large datasets.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
How do you scale Featuretools on large datasets?
You can install the optional Dask extension to run Deep Feature Synthesis across multiple cores or cluster nodes.
Can you create custom feature logic?
Yes, Featuretools allows you to define custom aggregation and transformation primitives in Python.
