摘要
To help government make emergency plans and citizens prepare protective measures, Forecasting PM2.5 concentration reliably is an important and achievable means for offering early warning. An improved forecasting model for PM2.5 concentration are proposed in this paper by constructing the novel correlation denoising and ensemble learning strategy. Empirical mode decomposition is first employed to construct the framework of ensemble learning strategy, which decomposes raw PM2.5 concentration into several intrinsic mode functions and one residual. Support vector machine coupled with correlation denoising is used to handle the severe-fluctuated intrinsic mode functions, while autoregressive integrated moving average is used to address the rest intrinsic mode functions and residual. The final forecasting results are obtained by summing the corresponding forecasted parts of support vector machine and autoregressive integrated moving average. Daily PM2.5 concentration and meteorological variables collected from three cities in central south China in 2018 and 2019 are used as case studies. The results of case studies answer why the proposed combination performs better than the other combinations in the same circumstances, and demonstrated that the proposed model benefited by correlation denoising and ensemble learning strategy improves forecasting accuracy and robustness over benchmark models.
学科与关键词
- Hourly PM2.5 concentration forecasting based on mode decomposition-recombination technique and ensemble learning approach in severe haze episodes of ChinaSun, Wei; Li, Zhaoqi*JOURNAL OF CLEANER PRODUCTION, 2020, 263: 121442.被引 42
- Application of decomposition-ensemble learning paradigm with phase space reconstruction for day-ahead PM2.5 concentration forecastingNiu, Mingfei; Gan, Kai; Sun, Shaolong*; Li, FengyingJournal of Environmental Management, 2017, 196: 110-118.被引 73
- A New Decomposition Ensemble Learning Approach with Intelligent Optimization for PM2.5 Concentration ForecastingXing, Guangyuan; Sun, Shaolong*; Guo, JueDISCRETE DYNAMICS IN NATURE AND SOCIETY, 2020, 2020: 6019826.被引 4
