alteryx/featuretools
Featuretools
Featuretools is an automated feature engineering library and data transformation framework written in Python. It automatically generates machine learning feature vectors from multi-table datasets by applying synthesis patterns to relational and timestamped data.
The system functions as a distributed feature synthesis engine, allowing the process of creating feature vectors to scale across multiple cores or clusters to handle large-scale datasets.
The library supports the synthesis of multi-table datasets, time series feature generation, and the creation of custom machine learning primitives when standard automated options are insufficient.
Features
- Automated Feature Engineering - Provides an automated framework for creating machine learning inputs from raw data to improve model accuracy.
- Deep Feature Synthesis - Implements deep feature synthesis to automatically generate feature matrices from relational data.
- Modular Primitives - Employs modular primitives as the architectural basis for transforming raw columns into engineered features.
- Custom Primitive Definitions - Provides interfaces for defining custom primitives to extend the automated feature synthesis logic.
- Python Machine Learning Libraries - Provides a Python-based toolkit for generating predictive feature vectors from raw timestamped data.
- Data Integration Frameworks - Provides a framework for extracting and transforming relational data into consolidated tables for predictive modeling.
- Relational Transformations - Applies relational transformations and aggregation patterns across multiple related tables to synthesize new features.
- Automated Feature Tables - Synthesizes single comprehensive tables containing all engineered features from multiple related dataframes.
- Schema-Driven Synthesis - Automatically discovers possible feature combinations by analyzing foreign key relationships within data tables.
- Distributed Computing - Utilizes distributed computing frameworks to scale feature synthesis across multiple nodes.
- Time-Window Aggregations - Implements time-window aggregations to generate temporal features and prevent data leakage.
- Large-Scale Data Computation - Processes large-scale data by distributing the synthesis of feature vectors across clusters.
- Large-Scale Feature Transformations - Executes feature engineering and transformations across massive datasets using distributed processing.
- ML Feature Stores - Generates a wide range of pre-computed feature vectors for use in machine learning models.
- Feature Generation - Automatically creates lag and time-based features from timestamped relational datasets.
- Feature Engineering - Library for automated feature engineering.
- Data Processing and Analysis - Automated feature engineering framework.
- Feature Engineering Tools - Automated feature engineering.
- Automated Machine Learning - Framework for automated feature engineering from relational data.
- Data Science Tooling - Automated feature engineering framework.
- Data Science Tools - Automated feature engineering framework.
