Air Quality Prediction Using a Novel Three-Stage Model Based on Time Series Decomposition

Air pollution control is one of the most important aspects of sustainable development in modern times, in which air quality prediction is the key in the process of governance. How to further improve the accuracy of air quality prediction and provide more powerful decision support for relevant departments is the main task of this paper. The air quality index (AQI) is a comprehensive indicator used to measure the air quality level. Taking AQI as the research object, this paper establishes a three-stage prediction model based on time series decomposition with improved particle swarm optimization—back propagation neural network (IPSO-BPNN), and makes empirical analysis of AQI in Wuhan City, China from 2014 to 2023. Specifically, at the first stage, the raw data is decomposed into season, trend, cycle and error subsequences by means of seasonal decomposition and empirical mode decomposition. At the second stage, the adaptive formula for the inertia weight in the particle swarm algorithm is newly defined by combining the sigmoid function with the scaling factor and control factor, and the improved particle swarm algorithm is used to optimize the initial weights and thresholds of BP neutral network. The cycle and error subsequences are predicted using this hybrid model. At the last stage, the four subsequences are recombined to obtain predictions of AQI. The empirical analysis results show that the MAPE of this hybrid prediction model reaches 0.04823, which is lower than the other two comparative models. The experimental results prove that the hybrid prediction model proposed in this paper can make an effective prediction.