Logo Lanfrica

MohamedYahmadi/Tunisian-Export-Ai

Domaine:

agriculture

Type de record:

software
Créateur:
Moh
Hôte:
AI-powered Tunisian agricultural export price prediction system using Java OOP, DJL, and Qwen 3 LLM. Features predictive analytics, market intelligence reports, and interactive dashboard. # 🇹🇳 Tunisian Agricultural Export AI System ## Project Overview An AI-powered Tunisian Agricultural Export Price Prediction and Market Intelligence System (Java 21). The application predicts future agricultural export prices, analyzes trends, and generates AI-driven market reports via ONNX Runtime (invoked through a Python prediction file) and LLM-based reporting, packaged in a Swing dashboard. ## Objectives - Predict future export prices for Tunisian products - Analyze market trends with statistical indicators - Generate intelligent market reports using LLMs - Demonstrate clean Java OOP and architectural patterns - Integrate modern AI libraries without heavyweight frameworks ## Key Features - **AI & Prediction**: ONNX Runtime-powered price forecasts (run from a Python inference file), batch predictions; confidence display and counts per product. - **Market Intelligence (LLM)**: Automated narrative reports; executive summary and detailed insights; French-oriented prompts (via `LLMReportService`). - **Dashboard & Analytics**: Swing UI with metrics (avg/max price, confidence, counts); table with filters; bar/line/historical charts; per-product navigation (Prev/Next); manual "Run Predictions" trigger; CSV export; report generation/download. - **Java OOP & Architecture**: Records/enums for domain, interfaces/abstractions for AI, annotations, stream-based data processing, exception handling, and modular package layout. ## Architecture ``` tn.isg.economics ├── Main.java # Application entry point (launches dashboard) ├── ai/ # ONNX + LLM services ├── dashboard/ # UI, charts, services, components ├── models/ # Domain records/enums ├── services/ # Business logic ├── exceptions/ # Custom exceptions └── Utils/ # Data mappers/loaders ``` Design patterns: MVC (dashboard), Strategy (charts), Factory (chart creation), observer-like updates, command-style UI a …