Bio-Inspired Optimization of Fuzzy Inference Rules for Air Quality Prediction in an Ensemble Framework

Abstract

This study presents a novel ensemble
learning approach designed to improve a fuzzy inference
system (FIS) for forecasting PM2.5 pollution levels. The
suggested model integrates the ensemble approach with
a FIS to enhance predictive accuracy. By developing a
collection of FIS, each trained on distinct subsets of the
data, this method utilizes model diversity to enhance
overall performance. Optimization algorithms are utilized
to refine the FIS parameters, thereby improving the
model’s predictive performance. The performance of the
optimized ensemble FIS is assessed through the
analysis of a real-world dataset concerning PM2.5
pollution levels. The findings demonstrate that the
suggested approach surpasses the conventional
ensemble algorithm, such as the commonly utilized
Random Forest, in terms of accuracy and robustness.
The optimized ensemble FIS presents a compelling
approach for accurate air quality forecasting, highlighting
its significance as an essential instrument for
environmental assessment public health.