Urban Air Quality Prediction Based on Space-Time Optimization LSTM Model

Abstract

Accompanying the rapid urbanization, many developing countries are suffering from serious air pollution problem. The demand for predicting future air quality is becoming increasingly more important to government’s policy-making and people’s decision making. In this paper, we predict the air quality of next hours for monitoring station, considering air quality data, meteorology data, and transit index data. Based on the domain knowledge about air pollution, we propose a long short-term memory network (LSTM)-based approach (entitled LSTM-STOM), which consists of a spatial optimization component and a temporal optimization method. We optimize the time window size to improve the prediction accuracy of the memory network. At the same time, considering air pollutants’ spatial correlations, the existing features with the wind direction features are integrated, so that the model can capture the interaction between different regions. Performance evaluation of LSTM-STOM method for hourly air quality index prediction is carried out during 2016–2018, in Hangzhou, and it is found that LSTM models efficiently deal with the complexities and is immensely effective in air quality forecasting. Comparing with the previous air pollution prediction models, we have relative accuracy improvements on short-term, long-term and sudden changes prediction, respectively.