Improving Signup Conversion Using ML and A/B Testing

This project simulates a real-world marketing experiment where machine learning is used to personalize call-to-action (CTA) messages and A/B testing is conducted to validate the effectiveness of the personalization.

📊 Dataset

A synthetic dataset of 10,000 users with the following columns:

• device_type, browser, traffic_source
• scroll_depth, time_on_page
• signed_up (target variable)

🧠 Machine Learning

• Random Forest Classifier predicts probability of user signup.
• One-hot encoding for categorical variables.
• Evaluated using classification metrics and ROC-AUC.

🎯 CTA Personalization

• If the model predicts a probability > 0.5 → “Start Now & Save 30%”
• Else → “Learn More About Us”

🧪 A/B Testing

• Users are randomly split into two groups:
  • A (Control): Fixed CTA.
  • B (Test): Personalized CTA.
• Signup outcomes are simulated.
• Two-proportion Z-test used to check statistical significance.

📈 Visualization

Bar chart showing conversion rates for both groups.

🛠️ How to Run

1. Install dependencies:

pip install pandas numpy scikit-learn statsmodels matplotlib

2. Run the script:

python ml_ab_test.py

📂 Files Included

• ml_ab_test.py: Main Python script
• user_sessions.csv: Sample dataset
• README.md: Project overview

🧠 Author

Created by an AI/ML and data science enthusiast. Feel free to modify and use this project in your own portfolio.