Predicting Air Quality Time Series with a Logical Dendritic Neuron Model

Air quality affects human health, and the prediction of air pollutant concentrations can help to improve air quality. For this purpose, a logical dendritic neuron model (LDNM), which has a dendritic structure, was used to predict air pollutant concentrations. To improve the prediction capability of LDNM, the gradient-based optimizer (GBO) algorithm was used to train LDNM. Considering the chaos of the air pollutant concentration, the time series were processed using phase space reconstruction (PSR). Time series of SO 2 and NO 2 concentrations were used in the experiments. The experimental results showed that LDNM outperformed the other methods in several evaluation metrics.