We explore the application of RNNs, using recurrent layers like LSTM, GRU and Bi-LSTM, in predicting air pollution levels, discuss the data preprocessing steps, evaluate the model’s performance, and compare it with other machine learning approaches. Through these experiments, we aim to assess the strengths and limitations of RNNs for air quality forecasting, while also highlighting the potential improvements that can be made to enhance prediction accuracy. Our approach will use RNNs and a sliding window algorithm in order to learn in an incremental manner. The models also have the possibility to change the number of input features during training without forgetting the model’s weights, this will be useful to test the scenario in which the number of input features grows with time.
(Code for the Big Data project)