AttriNeuLift 🧠📈

Attribution Uplift Engine – A neural uplift modeling approach to estimate user conversion likelihood by attributing the contribution of features.

🔍 Overview

AttriNeuLift is a machine learning pipeline aimed at solving the uplift modeling problem — predicting the incremental impact of features (or "touchpoints") on a user's decision to convert.

It does so by learning from user data with 12 numerical features, targeting a binary classification output: conversion vs. no conversion.

Uplift modeling is particularly useful in:

• Marketing: Understanding which factors actually influence conversions.
• Causal Inference: Modeling the true effect of features rather than correlations.
• Personalization: Targeting users with interventions that are likely to shift their behavior.

This project uses neural networks for uplift estimation with implementations of T-learners via the causalml library.

🧪 Key Components

• Criteo Dataset: Dataset containing anonymized user touchpoints. You can access the dataset here on Kaggle.
• Feature Attribution: Understand which features drive uplift.
• CausalML: Uses XGBTRegressor from CausalML for uplift modeling.
• Neural Network: Used MLPTRegressor from CausalML
• Saved the model for future use using Pickle
• Evaluation Metrics:
  • Conversion lift and Uplift Gain
  • Contribution of each variable to the final purchase decision

🛠 Technologies Used

• Python 3.12+
• Jupyter Notebooks
• Pandas, NumPy, Matplotlib, Seaborn
• scikit-learn
• causalml for uplift modeling
• PyTorch
• xgboost
• Pickle

📄 License

This project is licensed under the MIT License.