An Intelligent IoT-Cloud-Based Air Pollution Forecasting Model Using Univariate Time-Series Analysis

Area of Science:

• Environmental Science and Engineering
• Computer Science and Artificial Intelligence
• Data Science and Analytics

Background:

• Air pollution poses significant global health and environmental risks, exacerbated by industrial and vehicular emissions.
• Forecasting the Air Quality Index (AQI) is crucial for public health, but traditional models struggle with the volume of Internet of Things (IoT) data.
• Artificial intelligence (AI) and time-series models offer potential solutions for accurate AQI prediction.

Purpose of the Study:

• To assess the efficacy of an IoT-Cloud-based model for forecasting AQI under diverse meteorological conditions.
• To introduce and evaluate a novel Bayesian optimization-hybridized time-series (BO-HyTS) model for air pollution prediction.
• To compare the performance of the proposed BO-HyTS model against various classical, machine learning, and deep learning forecasting methods.

Main Methods:

• Developed a novel BO-HyTS model combining Seasonal Autoregressive Integrated Moving Average (SARIMA) and Long Short-Term Memory (LSTM), optimized via Bayesian optimization.
• Utilized Internet of Things (IoT) devices for real-time air quality data collection within a cloud environment.
• Employed five statistical evaluation metrics and a non-parametric Friedman test for rigorous model performance assessment.

Main Results:

• The proposed BO-HyTS model demonstrated superior performance, achieving Mean Squared Error (MSE) of 632.200, Root Mean Squared Error (RMSE) of 25.14, Median Absolute Error (Med AE) of 19.11, Maximum Error of 51.52, and Mean Absolute Error (MAE) of 20.49.
• BO-HyTS effectively captured both linear and nonlinear characteristics of time-series air pollution data, enhancing forecasting accuracy.
• Statistical significance tests confirmed the superior performance of BO-HyTS over competing AQI forecasting models.

Conclusions:

• The BO-HyTS model provides a highly accurate and efficient solution for AQI forecasting using IoT data.
• Study findings offer valuable insights into future AQI patterns in Indian states, informing public health policy.
• The proposed model can aid governments and organizations in proactive environmental management and policy development.