Smart Air Quality Prediction and Alert System for Android

Project Domain / Category:

AI / Mobile Apps

Abstract / Introduction:

"And the movement of the winds, and the submissive clouds between heaven and earth; certainly, in all these are signs (to recognize Allah) for the intelligent ones." [The Holy Quran: Para 2, Surah 2 Al-Baqarah, Verse 164]

Air pollution is a growing public health concern, making timely monitoring and prediction essential. This project "Smart Air Quality Prediction and Alert System for Android" introduces a machine learning–based mobile application that forecasts Air Quality Index (AQI) using historical AQI data and weather parameters like temperature, humidity, and wind speed.

The system applies predictive models to generate AQI forecasts and classifies them into health-based categories (Good, Moderate, Unhealthy, etc.). When poor air quality is detected, the app sends real-time alerts, helping users take preventive measures such as limiting outdoor activities or using protective gear.

Functional Requirements:

    User Management:

Support two types of users: Admin and Citizen

Provide registration and login pages for users.

Store credentials securely using Firebase Authentication.

    Admin Dashboard:

Manage historical AQI datasets.

Configure default AQI thresholds.

Update health recommendations for each AQI level.

Monitor user activity, manual data submissions, and alerts sent etc.

    Data Input (Automatic & Manual):

Automatic Input: Fetch real-time weather data (temperature, humidity, wind speed) and AQI data using public APIs, and refresh it automatically at regular intervals (e.g., every 60 minutes).

Manual Input: Citizens can manually input local data to get instant AQI predictions (even offline). Data should sync with Firebase database when internet is available.

    AQI Prediction:

Preprocess input data and pass it to the trained AI model.

Predict AQI value and classify it into AQI categories (Good, Moderate, Unhealthy, etc.).

Page 132 of 167

Display results within seconds along with its category, colour code and health recommendations.

   AQI Classification Levels:

The app should classify AQI into the following categories:

AQI

Range

Category

Color

Code

Health Impact

Recommended Action

   Alerts and Notifications:

Citizens should receive push notifications when AQI crosses unsafe thresholds.

Alerts should include recommended actions (e.g., "Avoid outdoor activities").

Admin can configure default AQI alert thresholds.

   History and Trends:

Store AQI history (daily, weekly, monthly) and visualize trends in graphical form.

Sync history with Firebase database for cloud backup.

   Offline Functionality:

Citizens should be able to view last fetched AQI and input data manually when offline.

Data should automatically sync when the device reconnects to the internet.

   Help and Support:

Provide FAQs, health guides, and tutorials to help users understand AQI and its impact.

Include a feedback option for reporting issues.

Workflow:

The basic workflow of this project is as follows.

Project Setup:

Install and configure PyCharm (for model development & training) and Android Studio (for app development).

Page 133 of 167

Configure Firebase for authentication and cloud database.

Dataset Preparation:

Collect and preprocess historical AQI and weather datasets

Handle missing values, normalize data, and split into training, validation, and test sets.

AI Model Development:

Train a Machine Learning model to predict AQI based on weather and historical data.

Evaluate and fine-tune the model for accuracy.

Convert the final model to TensorFlow Lite for Android deployment.

Android App Implementation:

Use Java/Kotlin to create the Android app.

Integrate the AI model for real-time AQI prediction based on live data or user input.

Implement functional requirements as given above.

Tools:

IDE: Android Studio & PyCharm

Programming Language: Java/Kotlin & Python

Databases: Firebase Real-Time/Cloud Fire-Store & SQLite/Room (for local storage)

Dataset: Kaggle AQI Dataset (https://www.kaggle.com/datasets/hajramohsin/pakistan-air-quality-pollutant-concentrations)

AI Model: Random Forest (https://scikit-learn.org/stable/modules/ensemble.html#forest) / XGBoost (https://xgboost.readthedocs.io/en/stable/)

Model Deployment: LiteRT i.e., TensorFlow Lite (https://ai.google.dev/edge/litert)

APIs: OpenWeatherMap API (https://openweathermap.org/api) / AQICN API (https://aqicn.org/api/) / OpenAQ API (https://openaq.org/)

Note:

           This is an Android-only development project, so tools like Flutter or Dart for cross-platform development are NOT allowed.

         VU will NOT pay for any software/library/toolkit/API used in this project.

Supervisor:

Name: Muhammad Imran Afzal

Email ID: imran.afzal@vu.edu.pk

MS Teams ID: imranafzal126@outlook.com