FEDOT

Overview

FEDOT is a code-based framework for automating machine-learning pipeline creation. It is best suited to developers and researchers who want automation they can steer across several data types and prediction tasks. Its breadth and adjustable workflow make it useful for experimentation, but it is not the fit for readers who need a graphical interface.

Key features

FEDOT supports binary and multiclass classification, regression, clustering, and univariate or multivariate time-series forecasting. Inputs can come from CSV files, pandas DataFrames, NumPy arrays, or time-series CSV data; the framework also works with tabular, text, and image data, including multimodal combinations. That range suits projects where data formats or tasks vary, although the workflow remains code-oriented.

The framework covers preprocessing, model selection, tuning, cross-validation, and serialization. Its preprocessing handles infinite and missing values, binary and non-binary categorical features, and extra spaces in categorical data, reducing some routine preparation. Cross-validation defaults to five folds; users can add metrics to the optimizer to address potential bias. That flexibility helps users shape evaluation, but does not make model assessment automatic or remove the need to choose suitable metrics.

Automation is adjustable: users can omit parameters for full automation or supply them to guide partial automation and manual pipeline composition. Presets include best_quality, fast_train, stable, auto, gpu, ts, and automl, with auto as the default. The choice is useful for balancing convenience with control, but taking advantage of that control calls for comfort with a code workflow.

FEDOT uses GOLEM to optimize and learn graph-based pipelines with meta-heuristic methods. Its models come mostly from scikit-learn, statsmodels, and Keras; the project also says it supports scikit-learn, CatBoost, and XGBoost and permits integration of custom libraries. This gives practitioners room to work with familiar tooling. GPU evaluation uses RAPIDS, but is limited to Ridge, Lasso, LogisticRegression, RandomForestClassifier, RandomForestRegressor, KMeans, and SVC, so the GPU preset is not a general acceleration guarantee across the model range.

Installation is through pip with pip install fedot; optional image, text-processing, and DNN dependencies are available with fedot[extra]. The API can also be called from a console without Python code, saving predictions as CSV files. That offers a command-line route for straightforward use, while the product's workflow interface is code rather than a graphical builder.

Pricing

FEDOT is free and open source. Its FEDOT plan costs nothing and includes an open-source AutoML framework under the BSD 3-Clause license. There is no free trial because the software is already free, and no paid tier or seat, quota, or renewal terms are part of this offer. The project is distributed under the 3-Clause BSD license for use in projects and research.

Because this is a self-hosted framework rather than a hosted service, users run it in their own environment. The free plan keeps the full framework available without a subscription, but users should expect to work with its code-oriented interface rather than a managed visual workflow.

Platforms

FEDOT supports Windows, Linux, and macOS, and is also categorized for API use and self-hosting. This makes it relevant to teams running models in their own environments across common desktop operating systems; it is not presented as a web-based application.

Who it's for

FEDOT is a strong fit for ML practitioners who want automated model selection and feature engineering while retaining the option to guide pipeline construction. Researchers and developers handling tabular, text, image, or time-series work can benefit from its task coverage and integration with established ML libraries. It is less suitable for people looking for a no-code visual workflow or for users whose GPU workload depends on models outside the supported RAPIDS subset.

Pros and cons

• Pros: Free, open-source distribution under BSD 3-Clause makes the framework usable in projects and research without a paid plan.
• Pros: Adjustable automation, graph-based pipeline optimization, and multiple presets give practitioners ways to move between automatic runs and guided composition.
• Pros: Support for varied tasks, multimodal data, and common input formats broadens its use across different modeling projects.
• Cons: The code workflow raises the entry bar for readers seeking a graphical interface.
• Cons: GPU evaluation covers only seven named model types, limiting the reach of GPU support.
• Cons: Users need to configure extra dependencies for image, text-processing, or DNN use when those are required.

Alternatives

AutoML Software is a useful place to compare other tools in the category.

LightAutoML is another free, open-source Python library, with an Apache License 2.0 and Linux, macOS, Windows, web, and self-hosted platforms. Choose it instead if that platform mix or license is a better fit.

AutoGluon is also a free open-source Python library, under Apache 2.0, for Linux, macOS, Windows, and self-hosted use. It is a comparable option when its licensing or platform profile better suits the project.

Amazon SageMaker Autopilot is a paid API and web option with a free trial and pay-as-you-go pricing, billed only for use with no minimum fees or upfront commitments. Choose it when a web-accessible service and usage-based billing are preferable to running a free framework yourself.

Auto-PyTorch is a free, self-hosted option for Linux developed by the AutoML Groups of the University of Freiburg and Hannover. It is an alternative when those platform and project credentials are more relevant.

BigML has a free tier with unlimited tasks and storage, a 16 MB maximum dataset size per task, two parallel tasks, and one user. Its caps matter for small, bounded use; choose it if a hosted/API-oriented option with those limits fits better than FEDOT's code framework.

JADBio offers a free Basic plan with one seat, three projects, 50 MB upload, 500 MB storage, one model export, and Standard Support SLA. It is worth considering when those project and storage limits fit the work.

EvalML is another free option for API, Linux, macOS, self-hosted, and Windows use.

FLAML is a free option for API, Linux, macOS, self-hosted, and Windows use.

Verdict

Choose FEDOT if you want a free, self-hosted AutoML framework that spans varied tasks and data types while letting you steer pipeline automation. Its strongest case is the combination of graph-based optimization and adjustable control; look elsewhere if a graphical workflow is essential or if your intended GPU models fall outside its supported subset.