Retailrocket Recommendation System & Shop Simulation

An end-to-end e-commerce application and analytical pipeline built as an engineering thesis project at Silesian University of Technology (Politechnika Śląska). The system processes historical e-commerce logs to generate market basket insights using the Apriori algorithm and serves real-time product recommendations within a simulated online store.

---

Architecture & Team Roles

The system uses a single PostgreSQL instance isolated into three database schemas to maintain a strict separation of concerns:

Schema	Ownership	Description	Tech Stack
historical	Data Engineering @LonelyLake	Raw data ingestion, cleaning, and preparation of the "basket" format.	DuckDB, Python
shop	Web Development @ElPollaco	Operational shop data (users, carts, simulated checkout transactions).	ASP.NET Core, EF Core
ml	Data Science @Blazejost	Association rules generated by the Apriori algorithm, exposed via API.	FastAPI, mlxtend

---

Project Structure

├── data/                # Raw and processed datasets (Local only, Git ignored)
│   ├── raw/             # Place downloaded Retailrocket CSVs here
│   └── processed/       # DuckDB analytical storage files
├── database/            # Docker Compose & Database Initialization scripts
├── data-pipeline/       # ETL processes (DuckDB analytical engine)
├── backend/             # E-commerce web backend (C# / .NET)
└── ml-service/          # Association rule mining engine & REST API (Python)

Quick Start (Local Database Deployment)

1. Configure Environment Variables:

   Navigate to the database directory and copy the template environment file:

   cd database
   cp .env.example .env

   (Optional: Open the newly created .env file and change the password if needed).

2. Start the Database:

   Spin up the PostgreSQL instance with all predefined schemas and tables:

   docker compose up -d

   The database will be automatically initialized using init.sql. You can connect via DBeaver or any other client using the credentials defined in your local .env file (default values):

   • Host: localhost
   • Port: 5559 (or whatever you set as DB_HOST_PORT in database/.env)
   • Database: retailrocket
   • User: admin
   • Password: admin (or your custom password from .env)

3. Dataset Setup (Local Only):

   • Ensure the data/raw/ and data/processed/ folders exist in the project root.
   • Download the raw Retailrocket CSV files from the team's Google Drive (link shared in chat) and place them inside data/raw/.

Git & Development Workflow

To ensure clean collaboration, the team follows the GitHub Flow: main branch is protected. No direct commits allowed.

Create a feature branch for your work: feature/your-feature-name.

Open a Pull Request (PR) to merge into main. At least one team member must review it.

Keep code and comments strictly in English.