> ML_LIBRARY // ECONML_v1.0
EconML
Microsoft Research / PyWhy — Microsoft Research and PyWhy toolkit for estimating Heterogeneous Treatment Effects via Double Machine Learning.
probabilistic-modellingv0.15.1BSD-3-Clausequalified
Model Training
Accelerators:
CPU
Distributed Training:No
Model Inference
Inference Accelerators:
CPU
Deployment Targets:server
What It Does
- +Double Machine Learning (DML) using arbitrary ML estimators (LightGBM, XGBoost, Random Forests)
- +Causal Forests and Generalized Random Forests (GRF) for personalized effect estimation
- +Dynamic treatment effects over time using panel data
- +Policy learning for optimal personalized decision-making rules
What It Does Not Do
- -Process unstructured audio or video streams
- -Perform automatic causal discovery from raw observational data without assumptions
- -Run on edge mobile browsers
>Suitable Work Types
- Personalized pricing determining which customer segments should receive discounts
- Medical treatment personalization discovering which patient subgroups benefit most from a drug
- Evaluating educational intervention impacts using high-dimensional covariate adjustments
>Unsuitable Work Types
- Simple linear regression with 2 variables where statsmodels is 10x faster and simpler
- Sub-millisecond high-frequency bidding algorithms
Data Residency Implications
Runs strictly locally in memory. Zero external network telemetry.
Security Considerations
Permissive BSD-3-Clause license. Enterprise ready.
Operational Profile & Known Limitations
Maturity:mature
Learning Curve:expert
Ops Complexity:moderate
Cost Tier:free-oss
> Known Limitations:
- Causal Forest fitting on hundreds of thousands of samples with large covariate sets is computationally demanding on CPU.
Associated Incident Patterns (Incidentpedia)
Enforce safeguards and monitoring to guard against these documented real-world failure modes:
> Primary Evidence & Benchmark Citations
EconML Documentationofficial-docs • >=0.14.0, <=0.15.x
2026-09-25HIGH