SpringerLink

Quantum-inspired neural networks for time-series air pollution prediction and control of the most polluted region in the world - Quantum Machine Intelligence

Researchers in the field of quantum computing are primarily focused on quantum-inspired algorithms, which function effectively on classical computers while incorporating quantum principles. A significant challenge in heavily polluted areas is accurately predicting air pollution levels. For our study, we selected New Delhi, specifically the Anand Vihar air pollution station, known for its high concentrations of PM $$_{2.5}$$ 2.5 . We proposed an approach to predict PM $$_{2.5}$$ 2.5 levels based on various pollutant and meteorological parameters. Our model is an improved version of the quantum temporal convolutional network (QTCN), enhancing the traditional quantum convolutional neural network (QCNN) model. To evaluate our model’s performance, we used several metrics, including mean squared error (MSE), root mean squared error (RMSE), mean absolute percentage error (MAPE), mean absolute error (MAE), and the coefficient of determination (R $$^2$$ 2 score). Our proposed model achieved MAE and MAPE values of ( $$59.031 \pm 1.23$$ 59.031 ± 1.23 ) and ( $$80.642 \pm 3.12$$ 80.642 ± 3.12 ), respectively. Additionally, the RMSE exhibited a reduction of ( $$32.493 \pm 1.8$$ 32.493 ± 1.8 ) % in comparison to the traditional QCNN framework, whereas the R $$^2$$ 2 score demonstrated an enhancement of ( $$14.86 \pm 1.0$$ 14.86 ± 1.0 ) %. The quantum-inspired model we have developed showcases its superior capabilities and demonstrates a significant advancement in forecasting air pollution levels, thus contributing valuable insights into environmental monitoring and public health. This research underscores the potential of using advanced computational techniques to address pressing challenges in air quality management, ultimately fostering a deeper understanding of the dynamics that govern atmospheric pollutants.