This is a tool for demand forecasting, regression and ensemble‐based dynamic pricing optimization; ML-driven user profiling with personalized recommendation systems for real‐time sales and engagement optimization.

The project was developed as a part of my real freelance work and published with the client's approval. For privacy purposes, all client data has been replaced with synthetic examples.

The mini-app is designed as PoC to help e-commerce platforms optimize product pricing and personalize user experiences.

It can predict future sales based on historical data, competitor pricing and seasonality using time series models, dynamically compute real-time optimal pricing. Profiling is done by clustering users based on behavior tracking (clicks, time on page, purchase history). Users are segmented for personalized recommendations using collaborative filtering techniques.

• config.py - configuration parameters for all modules
• data_generation.py - contains the DataGenerator class; synthetic data generation for sales, competitor pricing and user behavior
• demand_forecast.py - DemandForecaster class for time series demand forecasting using ARIMA models and grid search
• pricing_optimizer.py - PricingOptimizer class for dynamic pricing optimizer using regression and random forest models
• user_profiling.py - user segmentation using PCA and k-means clustering
• recommendation.py - collaborative filtering based recommendation system; implements the RecommendationEngine class that leverages cosine similarity
• utils.py - helper functions for logging, plotting, data I/O, etc.
• main.py - main script to run the complete pipeline

1. Clone:

   git clone git@github.com:avrtt/price-profiler.git
   cd price-profiler

2. Create a virtual environment and install dependencies:

   python -m venv venv
   source venv/bin/activate # Windows: venv\Scripts\activate
   pip install -r requirements.txt

3. Run the main script:

   python src/main.py

MIT