AeroCast AI: Jakarta Air Quality Prediction & Simulation Platform

AeroCast AI is an intelligent air quality platform combining real-time environmental monitoring, interactive scenario simulation, and XGBoost prediction models to provide data support for urban environmental decision-making.

Basic Project Info:

• Author/Maintainer: Arga Yura Danendra (Computer Science student at Diponegoro University)
• Source: GitHub
• Original Title: AeroCast AI - Real-Time Urban Air Quality Intelligence Platform
• Link: https://github.com/yurawawawawa/aerocast-air-quality-platform
• Release Time: June 3, 2026

Project Background & Motivation

Urban air pollution has become a major environmental challenge affecting public health. Jakarta, one of the largest cities in Southeast Asia, faces severe air quality issues. Traditional environmental monitoring methods often only provide historical data, lacking predictive capabilities and interactive analysis tools.

AeroCast AI was developed to demonstrate how machine learning can support environmental monitoring and decision-making, providing an intelligent solution for urban environmental management through accessible interactive tools.

Core Functions of AeroCast AI

Real-Time Air Quality Monitoring

Integrates WAQI (World Air Quality Index) API to obtain real-time air quality data for Jakarta. Displays current environmental conditions and pollutant concentrations, and uses trained models to predict current air quality status—helping users arrange daily activities.

Interactive Scenario Simulation

Provides interactive sliders for users to manually adjust pollutant and weather parameters, simulating different environmental scenarios. Users can observe changes in prediction results in real time, which is valuable for understanding how various factors affect air quality (e.g., PM2.5 concentration changes under specific wind speed and humidity).

ML Prediction Engine

Uses XGBoost classifier for multi-category air quality classification:

• Good (Baik): Satisfactory air quality, low pollution risk
• Moderate (Sedang): Acceptable, sensitive groups may be affected
• Unhealthy (Tidak Sehat): Sensitive groups' health may be impacted
• Very Unhealthy (Sangat Tidak Sehat): Health alert for all

Health Advice System

Generates context-related health advice based on predictions, providing preventive guidance for outdoor activities and respiratory safety—translating technical predictions into practical health information.

Technical Architecture & Implementation

Input Features

Air Pollutant Indicators: PM10, PM2.5, SO₂, CO, O₃, NO₂ Meteorological Parameters: Temperature, humidity, wind speed, rainfall, rainfall duration, solar radiation

Tech Stack

Technology	Purpose
Python	Core development language
Streamlit	Web application framework for quick data app interfaces
XGBoost	Machine learning model (efficient gradient boosting algorithm)
Scikit-Learn	ML tool library
Pandas	Data processing and analysis
NumPy	Numerical computation
WAQI API	Real-time air quality data source
Joblib	Model serialization and loading

System Module Design

Real-Time Monitoring Module

Obtains real-time air quality data from external environmental monitoring services and uses trained ML models for prediction—ensuring data timeliness and accuracy.

Environmental Scenario Simulator

Allows users to manually modify pollutant and weather conditions to analyze potential impacts on air quality. Valuable for environmental research and policy formulation.

Prediction Engine

Processes environmental feature data and generates air quality classification and confidence scores via the trained XGBoost model—core decision component of the platform.

Recommendation Engine

Generates health and activity advice based on predictions, converting technical outputs into user-friendly action guidelines.

Practical Application Value

AeroCast AI demonstrates ML's potential in environmental science:

• City Planners: Data-driven decision support tools
• Citizens: Timely health risk warnings
• Researchers: Scalable air quality analysis framework

Its open-source nature allows developers in other cities to adapt its architecture to build local air quality monitoring platforms—critical for solving global urban air pollution issues.

Technical Highlights & Insights

1. End-to-End ML Application: Covers full ML lifecycle (data acquisition, feature engineering, model training, deployment)
2. Real-Time Data Integration: Successfully integrates external API data to ensure prediction timeliness
3. Interactive Visualization: Uses Streamlit to quickly build professional data app interfaces
4. Multi-Class Classification: Applies gradient boosting algorithms to environmental science problems
5. Model Interpretability: Lets users understand factor impacts via scenario simulation

Conclusion

AeroCast AI is an excellent ML application case, transforming complex environmental science problems into interactive, understandable intelligent tools. As global urbanization accelerates and pollution worsens, such solutions (combining real-time data, prediction models, and user-friendly interfaces) will play an increasingly important role. For developers learning to apply ML to real-world problems, this project provides valuable references.