A PM2.5 Concentration Prediction Model Based on CART–BLS

With the development of urbanization, hourly PM2.5 concentration in air is constantly changing. In order to improve accuracy prediction, a prediction model based on Classification and Regression Tree (CART) Broad Learning System (BLS) was constructed. Firstly, CART algorithm used segment dataset hierarchical way obtain subset with similar characteristics. Secondly, BLS trained by using data each subset, validation error minimized adjusting window number mapping layer network. Finally, for leaf tree, global local path from root node are compared, smallest selected. The collected this paper come Chine Meteorological Historical Data website. We selected historical Huaita monitoring station Xuzhou city experimental analysis, which included pollutant content meteorological data. Experimental results show that effect CART–BLS better than RF, V-SVR, seasonal models.