DeepCausalMMM
DeepCausalMMM
Advanced Marketing Mix Modeling with Causal Inference and Deep Learning
DeepCausalMMM is a state-of-the-art Python package that combines deep learning with causal inference to understand the impact of marketing channels on business KPIs while learning causal relationships between channels through Directed Acyclic Graphs (DAGs).
pip install deepcausalmmm
# or latest from GitHub:
pip install git+https://github.com/adityapt/deepcausalmmm.git
Key Features
Config-Driven: Every setting configurable via config.py
GRU-Based Temporal Modeling: Captures complex time-varying effects
DAG Learning: Discovers causal relationships between channels (upper-triangular mask by default, or opt-in NOTEARS continuous optimisation — see DAG and NOTEARS structure learning)
Multi-Region Support: Handle geographic segmentation naturally
Robust Statistical Methods: Huber loss, comprehensive metrics
Production Ready: Battle-tested configurations and performance
What Makes It Special
Learnable Coefficient Bounds: Channel-specific, data-driven constraints
Data-Driven Seasonality: Automatic seasonal decomposition per region
Advanced Regularization: L1/L2, sparsity, coefficient-specific penalties
14+ Interactive Visualizations: Complete dashboard with business insights
Response Curves: Non-linear saturation analysis with Hill equations
Budget Optimization: Constrained optimization for optimal channel allocation
DMA-Level Contributions: True economic impact calculation
Perfect For
Marketing Mix Modeling: Understand true channel effectiveness
Attribution Analysis: Discover incremental impact of each channel
Budget Optimization: Data-driven media planning and allocation
Causal Discovery: Learn how channels influence each other
Multi-Touch Attribution: Beyond last-click attribution models