Analytical Cyclic Division of Dataset for an ANN-Type Model: A Case Study in Air Quality Prediction in Sub-tropical Area

This research intends to show how an analytical cyclic division of a dataset can improve ANN-type models in predicting future situation of different air pollutants in small-sized urban cities. Similar to other sub-tropical cities, the four seasons are not significant but climate characteristics of Macao obviously includes warm and cold seasons. These make it difficult to train the models well. Thus, an effective analytical way of cyclic division of the dataset for a seasonal LSTM modeling can improve the prediction of air quality with the meteorological data. We used data from 2016 to 2020 as input, and the model was trained on a 24-h basis, weekly oscillation frequency and finally grouped into 2 warm and cold seasons. A small-sized urban city was selected to demonstrate this study with 21 meteorological variables, and wavelet decomposition was used to clearly see the obvious oscillation and cycle patterns. The contributions include using LSTM for the prediction of time series with multivariate inputs in Macao and observing how the 2 cyclic division of the time series dataset of air pollutants and meteorological conditions look like. It is also intended to show, using some indicators, why the multivariate dataset should be divided according to the 2 cold/warm seasons. Finally, the result of predicting the concentrations of air pollutants is presented.