DATASHOPANALYTICS
Web application that uses machine learning to predict purchase behavior in e-commerce based on real Google Analytics data
The Problem
Marketing and UX teams in e-commerce lack predictive tools to identify users with high conversion probability in real-time. Without these capabilities, it is difficult to personalize experiences, optimize campaigns, and improve conversion rates based on navigation behavior patterns.
The Solution
Development of a web application with interactive dashboard that uses machine learning algorithms (XGBoost, LightGBM, Random Forest, SVM) to predict purchase probability. The system includes a real-time behavior simulator and generates automatic personalized marketing and UX recommendations.
Impact
Tool capable of identifying users with high conversion probability, allowing marketing and UX teams to optimize campaigns, personalize experiences, and improve conversion rates through navigation behavior simulations.
Category
Artificial Intelligence
Completion date
April 2025
Tech Stack
Highlights
- 89.5% accuracy with Optuna optimization
- Interactive dashboard with Streamlit
- Real-time behavior simulator
- Automatic recommendation system
- Exhaustive analysis of UCI dataset
- Complete preprocessing pipeline
Desktop View
Tablet View
Mobile View
Key Features
- Real-time purchase probability prediction
- Conversion metrics visualization and temporal analysis
- Automatic generation of marketing and UX strategies
- Persistent history with SQLite and SQLAlchemy
- Experimentation with multiple ML algorithms
- Personalized recommendation system
Next project
THECRITICALLENS