CausalML: A Python Package for Uplift Modeling and Causal Inference Empowered by Machine Learning Methods

• Activity Number: 324 - Causal Inference and Machine Learning in Practice: Challenges Across Industry
• Type: Topic Contributed
• Date/Time: Tuesday, August 9, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Zhenyu Zhao* and Totte Harinen
• Companies: Tencent and Toyota Research Institute
• Keywords: causal inference; uplift modeling; causal learning

Abstract

CausalML is a python package implementing various uplift modeling and causal inference methods. A key feature of this package is that it is focused on methods empowering causal inference with machine learning, also known as causal learning or causal machine learning. Another important feature is that the package is designed to process large-scale data typical for technology companies, similarly to established Python libraries for traditional machine learning methods. The goal of CausalML is to bridge the gap between theoretical work on methodology and practical applications by making a collection of methods in this field available in Python. This package was initially released in 2019, and it passed 650K downloads and 2700 stars on GitHub as of Jan 2022. The growing adoption of this package reflects the interest and needs of applying causal learning in broad business use scenarios.

In this paper, the package authors will introduce three aspects of this package: 1. Design and concept of this package 2. Core methods in this package 3. Industrial applications