Multi-factor PM2.5 concentration optimization prediction model based on decomposition and integration

With the rapid expansion of increased energy consumption, the issue of air pollution comes to be increasingly critical. It is essential to achieve accurate PM2.5 concentration prediction for people's health and lives. Therefore, a multi-factor PM2.5 concentration optimization prediction model based on circulatory system based optimization (CSBO), variational mode decomposition (VMD), gated recurrent unit optimized by quantile regression (QRGRU), mountain gazelle optimizer (MGO) and least square support vector machine (LSSVM), named CSBO-VMD-QRGRU-MGO-LSSVM, is proposed. Firstly, RFECV is utilized to discover the optimal feature subset with the strongest relationship with PM2.5 concentration. Secondly, variational mode decomposition optimized by circulatory system based optimization, named CSBO-VMD, is proposed. CSBO-VMD is utilized to decompose PM2.5 concentration adaptively into a restricted number of intrinsic mode functions (IMFs). Then, gated recurrent unit optimized by quantile regression, named QRGRU, and least squares support vector machine optimized by mountain gazelle optimizer, named MGO-LSSVM, are proposed. The decomposed components IMFs and optimal feature subsets are predicted by QRGRU and MGO-LSSVM to generate the integrated prediction results of QRGRU and MGO-LSSVM, respectively. Finally, the prediction results of QRGRU and MGO-LSSVM are assigned weights by the inverse root mean square error blending to obtain the final prediction results. Considering the geographical location, population density and pollution risk, PM2.5 concentration in Beijing, Shenyang, Xi'an and Urumqi are predicted to demonstrate the efficiency and universality of the proposed model. As demonstrated by the results, the proposed model has the greatest prediction precision and effectiveness.