Bi-LSTM based Air Pollution Risk Likelihood Prognostication using Ensemble Approach

In recent decades, many metropolitan regions throughout the world, including Indian states, have dealt with extreme air pollution and the health risks that are associated with it, rendering real time air quality and the risk likelihood predictions due to air quality critical for taking preventative and remedial actions. In this research, we have used a bi-directional LSTM (Long Short Term Memory) method to forecast the severity of air pollution. The dataset used is "Air Quality Data in India (2015 - 2020)". We demonstrated that an ensemble of two Bi-LSTM that model the short and long-term effects of pollution severity levels may greatly enhance predictions. Furthermore, weather information data was used in the modelling because it has been shown to improve risk likelihood forecast accuracy. The preliminary results for many stations are shown to illustrate that the model outperforms previous methodologies.