LightweightMMM 🦇 is a lightweight Bayesian Marketing Mix Modeling (MMM) library that allows users to easily train MMMs and obtain channel attribution information. The library also includes capabilities for optimizing media allocation as well as plotting common graphs in the field.
It is built in python3 and makes use of Numpyro and JAX.
We have kept JAX as part of the dependencies to install when installing LightweightMMM, however if you wish to install a different version of JAX or jaxlib for specific CUDA/CuDNN versions see https://github.com/google/jax#pip-installation for instructions on installing JAX. Otherwise our installation assumes a CPU setup.
The recommended way of installing lightweight_mmm is through PyPi:
pip install --upgrade pip
pip install lightweight_mmm
If you want to use the most recent and slightly less stable version you can install it from github:
pip install --upgrade git+https://github.com/google/lightweight_mmm.git
Issue tracker: https://github.com/google/lightweight_mmm/issues
Source code: https://github.com/google/lightweight_mmm/tree/main
LightweightMMM is licensed under the Apache 2.0 License.