Analysis and Prediction of Air Pollutant Indices using Bidirectional-Convolutional LSTMs

Abstract

Air pollution is a crucial issue that affects people’s health and simultaneously the planet’s future. Forecasting is a key role for preventing and preparing the public to combat it. Meteorological forecasting is a key approach to the problem that uses numerous practices. Some of the most well-known are Regressions analysis and Neural Networks. Both use historical datasets in order to determine forthcoming pollution levels. For Neural Networks the LSTMs, Conv1D and Transformers are some of the best tools to tackle such important issues. This paper presents an alternative solution for time series prediction for air pollution using a BiLSTM-Conv1D neural network which rivals other models on their performance.