Author(s)

Dongpo Cheng

Corresponding Author

Dongpo Cheng

Abstract

Since thetime series data often arrives in the form of data flow, and the potential distribution and the trend of PM2.5 the data change continuously with time, in this case, the online learning model with incremental update capability is more suitable for processing these non-stationary time series prediction problems. In this paper, the Online Recurrent Extreme Learning Machine (OR-ELM) is applied to the concentration prediction field for the first time. Compared with the existing PM2.5 concentration prediction model, the model improves the accuracy of PM2.5 concentration prediction by online learning. In order to verify the effect and superiority of the proposed model, this paper selects the hourly data of PM2.5 concentration in Beijing, Shanghai, Xi'an and Chongqing, and then conducts empirical research, and finally compares with benchmark models such as random forest (RF) ,gradient lifting tree (GBDT) and multi-layer perceptron model (MLP) in predictive performance . This model can provide a reference system for air quality warning due to its excellent predictive performance.

Keywords

PM2.5,Time series predicting, Online learning ,Online Recurrent Extreme Learning Machine（OR-ELM）