SW2627-Data-Product-Development-YTOR

YTOR is an operational dashboard designed to identify seller behaviors in an e-commerce marketplace that consistently reduce customer trust over time. It correlates seller fulfillment metrics (misleading item returns, late dispatches, post-sale order cancellations) with customer review sentiment decay to calculate a dynamic Seller Trust Index (0-100) and trigger automated operational enforcement actions.

---

🛠️ Tech Stack & Architecture

• Language: Python 3.11+
• Data Manipulation: Pandas, NumPy
• Database & Querying Layer: SQL / SQLite
• Interactive Dashboard: Streamlit, Plotly
• Styling: Modern dark theme glassmorphism CSS
• CI/CD Pipeline: GitHub Actions (Data seeding & pytest automated validation)

---

📊 Features & Core Concepts

1. Seller Trust Index (0-100)

The Trust Score evaluates sellers dynamically over configurable sliding windows (30/60/90/180 days):

• Misleading Return Penalty: Deducts points for returns categorized under Misleading Description or Defective Product.
• Late Dispatch Penalty: Penalizes high delay ratios against promised delivery dates.
• Cancellation Penalty: Severe penalties for post-payment seller cancellations due to stockouts.
• Sentiment Decay Penalty: Monitors drops in moving-average review sentiment scores.
• Support Latency Factor: Penalizes long dispute resolution lead times.

2. Risk Tiers

• Critical Risk ($T < 50$): Immediate visibility restriction and catalog audit required.
• Moderate Risk ($50 \le T < 70$): Formal operational warning issued with mandatory support SLA.
• Watchlist ($70 \le T < 85$): Fulfillment audit and review sentinel monitoring.
• Low Risk / Healthy ($T \ge 85$): High trust score, eligible for Top Seller badge.

3. Interactive Streamlit Dashboard

• Executive Trust Overview: Marketplace-wide KPI cards, risk tier donut chart, and primary trust penalty drivers.
• Behavior & Sentiment Deep-Dive: Scatter plot of Misleading Returns vs. Negative Review Sentiment, correlation matrix heatmap, and individual seller historical trajectories.
• Operations Action Center: Recommended enforcement actions table with CSV export.
• SQLite Data Explorer: Raw database table views and custom SQL query workbench.

---

🚀 Quickstart Guide

1. Clone & Install Dependencies

git clone https://github.com/your-username/ytor.git
cd YTOR
pip install -r requirements.txt

2. Seed SQLite Database

python -m src.generator

3. Launch Streamlit Dashboard

streamlit run app.py

4. Run Pytest Suite

python -m pytest tests/

---

📁 Repository Structure

YTOR/
├── .github/
│   └── workflows/
│       └── pipeline.yml         # GitHub Actions workflow for CI/CD
├── data/
│   ├── schema.sql               # SQLite schema definition
│   └── ytor.db                  # Generated SQLite database
├── src/
│   ├── __init__.py
│   ├── db.py                    # SQLite connection & query helper module
│   ├── generator.py             # Realistic marketplace data generator
│   ├── trust_model.py           # Trust Index math & risk tier algorithms
│   └── analytics.py             # Pandas/NumPy analytics & aggregations
├── tests/
│   ├── test_db.py               # Unit tests for database & schema
│   └── test_analytics.py        # Unit tests for trust engine & metrics
├── .streamlit/
│   └── config.toml              # Streamlit layout & color theme config
├── app.py                       # Main Streamlit web application
├── styles.css                   # Custom CSS styling
├── requirements.txt             # Dependency requirements
└── README.md                    # Project documentation