Estimating Fine Dust Concentration with LSTM
This study suggested an effective method for interpolating missing values of fine dust concentration data measured in monitoring stations used in fine dust analysis studies conducted to solve fine dust. At this time, Long Short Term Memory, one of the Recurrent Neural Network models that considers the time dependencies of data, was used. For the input data of the model, the concentration data of the neighbor stations, the neighbor ASOS, and Getis-Ord Gi*, which can imply the spatial relationship of the stations, were used. As a result of the study, the accuracy of interpolation of missing values of the model presented in the study was higher than that of the existing spatial interpolation method.