🌫️ PM2.5 Forecast App (Website Admin)

PM2.5 Forecast is a mobile application developed using Flutter (Dart). It displays real-time information on PM2.5 levels, air quality, and weather conditions. The app also provides future PM2.5 forecasts using an AI-powered LSTM (Long Short-Term Memory) model.

This project was developed as part of a university assignment in Year 3, Semester 2.

---

🚀 Key Features

• 🧭 Location-Based Station Detection – Automatically find the nearest air quality monitoring station
• 📊 PM2.5 Forecasting (LSTM) – Hourly PM2.5 prediction using deep learning
• 🌦️ Real-Time Weather – Display temperature, humidity, rainfall, and AQI (Air Quality Index)
• 🗺️ Interactive Map View – View all monitoring stations on a map
• 🔍 Station Search – Search stations by province or name
• 💡 Health Advice – Provide health recommendations based on pollution levels
• 🧠 AI Model Integration – LSTM-based forecasting from historical data
• 📦 Local Storage – Store last visited station using SharedPreferences

---

⚙️ System Architecture

The PM2.5 Forecast system is divided into 4 major components, each responsible for a specific part of the application:

Component	Description
📱 Flutter App	Cross-platform mobile app developed in Flutter and Dart
🌐 Node.js Backend	RESTful API handling real-time and forecast data
🧠 AI Engine (Python)	LSTM model for PM2.5 forecasting
📡 External Data Sources	Data from meteorological and environmental authorities (e.g., Thai Meteorological Dept, PCD, GISTDA)

All components are loosely coupled and communicate via REST APIs.

---

🧰 Tech Stack

💻 Frontend (Flutter)

• Flutter + Dart
• http, shared_preferences, google_maps_flutter, intl
• flutter_svg, bottom_navigation_bar
• State management with FutureBuilder
• GPS and real-time graph plotting
• UI designed with Figma

🌐 Backend (API)

• Node.js + Express
• RESTful APIs:
  • PM2.5 levels per station
  • Health recommendation endpoints
  • Forecast data endpoints
• MySQL or PostgreSQL database

🧠 AI Model

• Python with TensorFlow / Keras
• Model: Long Short-Term Memory (LSTM)
• Techniques: Data Pre-processing, Normalization, Sliding Window
• Evaluation Metrics:
  • Mean Squared Error (MSE)
  • Root Mean Squared Error (RMSE)
  • Mean Absolute Error (MAE)
  • Mean Absolute Percentage Error (MAPE)

📖 Theoretical Foundations

• The LSTM (Long Short-Term Memory) model is used to forecast PM2.5, leveraging its ability to capture long-term time series patterns.
• Data collected from multiple sources:
  • Thai Meteorological Department
  • Pollution Control Department
  • Traffic reports (e.g., Google Traffic)
  • GISTDA spatial data

Model training is based on historical data and evaluated using MSE, RMSE, MAE, and MAPE

🎓 Academic Context

This project demonstrates skills in:

• Mobile app development using Flutter
• Full-stack system integration with REST APIs
• Sensor and weather data processing
• Forecasting using LSTM deep learning
• Data fusion and preprocessing techniques
• UI/UX design with Figma
• Real-time interactive visualization

🧪 Testing & Tools

• ✅ Tested on Emulator & Physical Android Devices
• 🛠️ Visual Studio Code, Android Studio
• 📊 Power BI for data analysis and dashboarding
• 📐 Draw.io for diagrams and system design
• 💾 MySQL Server for relational data management

---