EvalML is a free AutoML library for building, optimizing and evaluating machine-learning pipelines with domain-specific objective functions. It can construct pipelines with preprocessing, feature engineering, feature selection and multiple modeling techniques, and includes tools for examining models. Listed automation features include data-quality checks and cross-validation. Users can select standard objectives such as mean squared error, cross entropy and area under the ROC curve, or define custom objectives. EvalML can be combined with Featuretools and Compose for end-to-end supervised machine-learning solutions. Its time-series functionality uses Prophet to predict future values from past ones; that support is still being developed. Tutorials cover fraud prediction, lead scoring, cost-benefit objectives and text data. Installation is available through PyPI, conda-forge or source, with Python 3.9–3.11 supported on the current installation page. Optional dependencies provide XGBoost, CatBoost and plotting support. It supports Linux, macOS, Windows, API use and self-hosted setups, but dependency configuration can vary by platform: Mac use requires OpenMP for LightGBM, and Apple M1 dependency support is incomplete.
Who it is for
EvalML suits people building supervised machine-learning pipelines in code who want automated pipeline construction and custom objectives. Its tutorials also address fraud prediction, lead scoring, cost-benefit objectives and text data.
What is good
- Free AutoML library with no trial requirement.
- Automates pipeline construction and optimization.
- Supports standard and custom objectives.
- Includes model inspection tools.
- Can work with Featuretools and Compose.
What to know first
- Time-series support is still being developed.
- Mac requires OpenMP for LightGBM.
- Apple M1 dependency support is incomplete.
- Some Windows dependencies may need conda installation.
Verdict
EvalML provides pipeline automation, configurable objectives and model inspection without a listed price. Platform-specific dependency requirements and developing time-series support are worth considering before installation.
Compared on AutoML software
- Free plan
- Yes
- Feature engineering
- Yes
- Automated model selection
- Yes
- Model explainability
- Yes
- Workflow interface
- code
- Hosting model
- self_hosted
