Prediction of PM2.5 Concentration Based on the LSTM-TSLightGBM Variable Weight Combination Model

Abstract

PM2.5 is one of the main pollutants that cause air pollution, and high concentrations of PM2.5 seriously threaten human health. Therefore, an accurate prediction of PM2.5 concentration has great practical significance for air quality detection, air pollution restoration, and human health. This paper uses the historical air quality concentration data and meteorological data of the Beijing Olympic Sports Center as the research object. This paper establishes a long short-term memory (LSTM) model with a time window size of 12, establishes a T-shape light gradient boosting machine (TSLightGBM) model that uses all information in the time window as the next period of prediction input, and establishes a LSTM-TSLightGBM model pair based on an optimal weighted combination method. PM2.5 hourly concentration is predicted. The prediction results on the test set show that the mean squared error (MAE), root mean squared error (RMSE), and symmetric mean absolute percentage error (SMAPE) of the LSTM-TSLightGBM model are 11.873, 22.516, and 19.540%, respectively. Compared with LSTM, TSLightGBM, the recurrent neural network (RNN), and other models, the LSTM-TSLightGBM model has a lower MAE, RMSE, and SMAPE, and higher prediction accuracy for PM2.5 and better goodness-of-fit.

A Spatial–Temporal Causal Convolution Network Framework for Accurate and Fine-Grained PM2.5 Concentration Prediction

“…Apart from fine-grained air pollutant concentration prediction, RNN-based models and graph-based models also show strong ability in intelligent transportation system (ITS), recommender system (RS), stocking market and so on. However, the natural defects in these two kinds of models should not be neglected, especially in fine-grained PM 2.5 concentration prediction: (1) Gradient disappearance and gradient explosion, time consumption and large memory requirement [ 64 ] limit the application of RNN-based models, such as LSTM [ 27 , 28 ] and GRU [ 32 , 33 ]; (2) Unstable factors in graph-based models, such as the Markov hypothesis [ 37 ], and difficulty in capturing the dynamic edge relationship between stable nodes [ 65 ], make it difficult for graph-based models to effectively extract spatial and temporal features in fine-grained PM 2.5 concentration prediction especially in a wide research area (i.e., city-level and regional level prediction) [ 65 , 66 ]; (3) The state-of-the-art models (i.e., ConvTrans [ 67 ], Informer [ 68 ], FC-GAGA [ 69 ], MAGCN [ 70 ] and Multi-STGCnet [ 71 ], etc.) which focus on fine-grained time series prediction behave not so optimistically if these models are directly employed into fine-grained PM 2.5 concentration prediction.…”

“…The RNN proposed by Hopfield can model time series data and extract the time dependence of context [ 25 ]. Subsequently, the RNN variant model LSTM [ 26 , 27 , 28 , 29 , 30 , 31 ] and GRU [ 32 , 33 , 34 ] proposed to solve the short-term memory problem caused by the disappearance of the RNN gradient. Zhang et al apply the ConvLSTM model to model the data of sky stations and daily aerosol optical thickness to predict the daily spatial distribution of PM 2.5 concentration [ 35 ].…”

Microphysical Characteristics and Environmental Isotope Effects of the Micro-Droplet Groups under the Action of Acoustic Waves

“…where δ is the isotopic composition (‰), R sa represents the isotopic ratio of the tested samples, and R sm is the isotopic ratio of a standard sample. The isotope ratio was calculated using Equation (6).…”

“…Atmospheric aerosol particles are of great concern because of their adverse effects on human health and the environment [1][2][3][4][5][6][7][8]. Especially for PM 2.5 and PM 10 , the removal of fine particles has attracted the extensive attention of scholars [9][10][11].…”

Fine-Grained Individual Air Quality Index (IAQI) Prediction Based on Spatial-Temporal Causal Convolution Network: A Case Study of Shanghai

“…The key to IAQI modeling is to extract time dependence and spatial correlation from data. Current methods for extracting time-dependent features (especially features over a long time span) include recurrent neural network (RNN) [8,9], gate recurrent unit (GRU) [10][11][12] and long-short term memory (LSTM) [13][14][15][16][17]. However, these RNN-based sequential approaches suffer from time-consuming iterative propagation, gradient explosion and vanishing problems [18].…”