Research and application of a novel selective stacking ensemble model based on error compensation and parameter...

Area of Science:

• Environmental Science
• Data Science
• Artificial Intelligence

Background:

• Industrialization is a major driver of declining air quality globally.
• Accurate Air Quality Index (AQI) prediction is crucial for public health and environmental monitoring.
• Existing AQI prediction models often lack sufficient accuracy to address complex atmospheric conditions.

Purpose of the Study:

• To develop an advanced methodology for improving the accuracy of Air Quality Index (AQI) prediction.
• To integrate stacking ensemble techniques with error correction for enhanced predictive performance.
• To optimize model parameters using the reptile search algorithm (RSA) for robust AQI forecasting.

Main Methods:

• Collected and utilized four distinct regional AQI datasets comprising 34,864 data samples.
• Performed cross-validation on ten common single predictive models to establish baseline performance.
• Selected five top-performing models for stacking ensemble based on evaluation indices.
• Employed the reptile search algorithm (RSA) for hyperparameter optimization.

Main Results:

• The proposed stacking ensemble and error correction model demonstrated a significant improvement in AQI prediction accuracy.
• Achieved approximately a 10% increase in prediction accuracy compared to conventional AQI forecasting models.
• The reptile search algorithm effectively optimized the parameters of the ensemble model.

Conclusions:

• The novel methodology offers a more scientifically grounded and accurate approach to AQI prediction.
• This enhanced prediction capability can aid in more effective air pollution management strategies.
• The study highlights the potential of combining ensemble methods, error correction, and metaheuristic optimization for environmental monitoring.