GameEdge Intelligence: AI-Powered Sports Betting Analytics Platform

GameEdge Intelligence is a comprehensive sentiment analysis and customer segmentation platform for sports betting and fantasy sports, built with enterprise-grade architecture and advanced ML capabilities. It combines BERT-based transformers with traditional ML fallback, RFM analysis, clustering, churn prediction, and real-time analytics over large-scale sports and transaction datasets—1.6 million pre-labeled tweets, 525k+ transaction records, and 500k+ sports matches with odds data power the training and demo pipelines.

Full-stack architecture

The platform splits cleanly between a Python FastAPI backend with async ML operations and a Next.js 14 frontend using App Router, TypeScript, Tailwind CSS, and Shadcn/ui. PostgreSQL with SQLAlchemy ORM stores customer profiles, segments, and analytics aggregates. WebSockets deliver live sentiment feeds to the dashboard. Docker Compose orchestrates development and production-like environments with environment-specific configuration files.

Backend module layout

• app/api/ — REST endpoints for sentiment, segmentation, analytics, and data-pipeline control
• app/core/ — Configuration, security settings, and database session management
• app/models/ — SQLAlchemy ORM models for customers, transactions, and sentiment events
• app/services/ — Business logic separating HTTP transport from ML and analytics rules
• app/ml/ — Model training, inference pipelines, and monitoring hooks

ML pipeline design

• Multi-model sentiment — BERT transformers via Hugging Face with scikit-learn fallback for resilience
• Customer segmentation — RFM analysis, clustering algorithms, and churn prediction models
• Automated preprocessing — Feature engineering, train/validation splits, and model versioning
• Real-time inference — Async FastAPI handlers with thread-pool or worker isolation for CPU/GPU-bound steps
• Drift and performance monitoring — Hooks for retraining triggers when sentiment or churn distributions shift

Data pipeline architecture

scripts/manage_data_pipeline.py provides CLI control over pipeline status, synthetic data generation, Kaggle dataset downloads, and per-dataset transforms. Data pipeline APIs expose GET /api/v1/data-pipeline/status, POST run/synthetic/transform endpoints, and pipeline metadata for operators. Synthetic generation supports configurable user counts for testing without external dependencies; real dataset ingestion requires Kaggle credentials and respects licensing constraints.

Real-time analytics

WebSocket /ws/live-sentiment streams sentiment events to the frontend for live monitoring dashboards. Recharts and D3.js render time-series sentiment trends, segment distributions, and churn risk heatmaps. The dark-theme UI targets sports betting operators who need high-contrast operational views during live events.

Security and deployment

• JWT_SECRET for authenticated API access in production configurations
• DATABASE_URL and REDIS_URL environment variables for managed service wiring
• ML_MODEL_PATH for versioned artifact deployment separate from application code
• docker-compose.prod.yml for production container orchestration
• CI/CD workflows under .github/ for automated test and build gates

API surface summary

• POST /api/v1/sentiment/analyze — Text sentiment scoring with model selection
• GET /api/v1/customers/segments — Retrieve customer segment definitions and metrics
• POST /api/v1/customers/segment — Create or update segmentation rules
• GET /api/v1/analytics/dashboard — Aggregated KPIs for executive and operator views
• GET /api/v1/predictions/churn — Churn probability scores with feature contributions

Documentation and extension

The repository includes docs for API reference, data pipeline operations, ML pipeline guides, deployment, and contributing. Extension paths include federated learning for multi-book operators, real-time odds integration APIs, responsible-gaming compliance modules, and feature stores for cross-sport customer intelligence.