Aquaculture AI 🐟

An intelligent aquaculture farm management system powered by machine learning.

Features

• Water Quality Engine (WQE): Analyze, classify, and predict water quality
• Anomaly Detection: Detect sensor failures and unusual patterns
• Time-Series Prediction: Forecast pH, DO, temperature, and ammonia
• Actionable Recommendations: Get expert-system recommendations
• REST API: Deploy as a service

Quick Start

Installation

# Clone or navigate to the project
cd aquaculture_ai

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Linux/Mac
# or
.\venv\Scripts\activate  # Windows

# Install dependencies
pip install -r requirements.txt

Basic Usage

from src.engines.water_quality import WaterQualityEngine

# Initialize engine
engine = WaterQualityEngine()

# Generate synthetic data and train all models
results = engine.train_all(n_synthetic=5000)

# Analyze current water quality
analysis = engine.analyze({
    "pH": 7.5,
    "DO": 5.2,
    "temperature": 28.5,
    "ammonia": 0.3
})

print(f"WQI Score: {analysis['wqi_score']}")
print(f"Quality: {analysis['status']['quality_class']}")
print(f"Stress Level: {analysis['status']['stress_level']}")
print(f"Alerts: {len(analysis['alerts'])}")
print(f"Recommendations: {analysis['recommendations']}")

Run API Server

from src.engines.water_quality import WaterQualityEngine
from src.api.endpoints import run_server

# Initialize and train engine
engine = WaterQualityEngine()
engine.train_all()

# Start API server
run_server(host="0.0.0.0", port=8000, engine=engine)

Then access:

• API Docs: http://localhost:8000/docs
• Health Check: http://localhost:8000/health

Project Structure

aquaculture_ai/
├── src/
│   ├── config.py              # Configuration & thresholds
│   ├── data/
│   │   ├── loader.py          # Data loading
│   │   ├── cleaner.py         # Data cleaning
│   │   ├── preprocessor.py    # Feature engineering
│   │   └── schema.py          # Data schema validation
│   ├── engines/
│   │   └── water_quality.py   # Main Water Quality Engine
│   ├── models/
│   │   ├── classifier.py      # XGBoost classifier
│   │   ├── lstm_predictor.py  # LSTM time-series
│   │   └── anomaly_detector.py # Isolation Forest
│   ├── utils/
│   │   ├── metrics.py         # Evaluation metrics
│   │   ├── visualization.py   # Plotting
│   │   └── alerts.py          # Alert system
│   └── api/
│       └── endpoints.py       # FastAPI endpoints
├── notebooks/                  # Training notebooks
├── data/                       # Data storage
├── saved_models/              # Trained models
└── tests/                     # Unit tests

Water Quality Engine

The WQE provides three main models:

1. Classifier

Classifies water quality into categories:

• Excellent (WQI > 85)
• Good (WQI 70-85)
• Moderate (WQI 55-70)
• Poor (WQI 40-55)
• Dangerous (WQI < 40)

2. LSTM Predictor

Predicts next 6-24 hours of:

• pH
• Dissolved Oxygen (DO)
• Temperature
• Ammonia

3. Anomaly Detector

Detects:

• Sensor failures
• Sudden spikes
• Unusual patterns

Configuration

Key thresholds in config.py:

Parameter	Safe Range	Optimal
pH	6.0 - 9.0	7.5
DO	4.0+ mg/L	6.0 mg/L
Temperature	24-32°C	26-30°C
Ammonia	< 0.5 mg/L	< 0.02 mg/L

API Endpoints

Endpoint	Method	Description
/health	GET	Health check
/analyze	POST	Analyze water quality
/predict	POST	Predict future values
/train	POST	Train models
/alerts	GET	Get active alerts
/thresholds	GET	Get thresholds

Future Engines

This project is designed to support additional engines:

• Growth Prediction Engine
• Survival Risk Engine
• Feeding Efficiency Engine
• Disease & Medicine Engine
• Yield Prediction Engine
• Root Cause & Recommendation Engine

License

MIT License