mckinsey/causalnex

A Python library that helps data scientists to infer causation rather than observing correlation.

What it solves

CausalNex simplifies the process of causal reasoning and "what-if" analysis. It addresses the problem where data scientists often need multiple libraries to learn causal structures, incorporate domain expertise, and estimate the effects of interventions using data.

How it works

The toolkit uses Bayesian Networks to identify causal relationships in data, moving beyond simple pattern recognition and correlation. It allows users to learn causal structures from data using state-of-the-art methods, manually augment those relationships with domain knowledge, and fit probability distributions to the network. Once the model is built, users can perform counterfactual analysis and use Do-calculus to analyze the impact of interventions.

Who it’s for

Data scientists and analysts who want to move from correlation-based machine learning to causal inference and identify the right interventions for a business or technical problem.

Highlights

  • own state-of-the-art structure learning methods to understand conditional dependencies
  • Ability to encode and augment domain expertise directly into the graph model
  • Support for Do-calculus to analyze the impact of interventions
  • Built-in visualization tools to simplify the understanding of causality
  • Tools for fitting probability distributions and evaluating model quality with statistical checks

Related

  • Project
    erdogant/bnlearn

    bnlearn is a Python library for learning, parameterising, and querying Bayesian networks. It supports structure discovery (various scores and search methods), CPT estimation, causal inference (do‑calculus), synthetic data generation, and visualisation, all wrapped in a simple API.

  • Project
    cdt15/lingam

    A Python library for causal discovery that estimates linear non-Gaussian acyclic models to determine the causal relationships between variables.

  • Project
    pgmpy/pgmpy

    A Python library for causal and probabilistic reasoning using graphical models, providing tools for structure learning, parameter estimation, and causal inference.

  • Project
    uber/causalml

    A Python package for uplift modeling and causal inference that estimates the causal impact of interventions on individual users to optimize targeting and personalization.

  • Project
    py-why/dowhy

    DoWhy is a Python library for end‑to‑end causal inference. It lets you define a causal graph, automatically identifies the causal effect, estimates it with a variety of statistical methods, and runs refutation tests to check robustness. The library also supports graphical causal models for root‑cause analysis, counterfactuals, and interventional sampling. Install via `pip install dowhy` and follow the four‑step workflow (model → identify → estimate → refute).