Utilizing Deep Learning techniques for a multi-objective pollution forecasting model to enhance smart city sustainability

Accurate pollutant forecasting serves as a crucial component in air quality monitoring and control in a smart city. Traditional pollutant forecasting models such as statistical, and machine learning models follow a single-input, single-output learning approach, which fails to identify multi-output dependencies. In addition to that the traditional forecasting model ignores the data missing values and spatial dependencies of nearby locations, which can affect air pollutant concentration levels. The research work adopted a multiple imputation technique, which imputes the missing values using spatial attributes. A multi-objective air pollutant forecasting approach is developed, combining advanced data imputation with multi-input and multi-output (MIMO) multi-step forecasting capabilities. In addition to that the proposed multi-objective based Multi-output Bidirectional Long Short-Term Memory (MBLSTM) model used Bidirectional Long Short-Term Memory (BiLSTM) with the MIMO approach. Overall, the proposed MBLSTM model provides improved multiple pollutants forecasting results in a single run.