Development of a Novel ANN-Based Predictive Model for Multi-Site ILCR Estimation Using Weather Parameters and PM2.5

This study introduces a novel artificial neural network (ANN)-based methodology for predicting the Incremental Lifetime Cancer Risk (ILCR) in urban environments, leveraging weather parameters and PM_2.5 concentrations. The innovative approach addresses the limitations of conventional ANN models, enabling superior performance and applicability across diverse geographical locations. The model incorporates a unique method to preprocess wind direction and speed into a singular representative factor, enhancing its adaptability and generalizability to different sites.

To validate the proposed methodology, one year of weather data and polycyclic aromatic hydrocarbons (PAHs) data were collected from two distinct sites in India. PAHs were analyzed using gas chromatography-mass spectrometry (GC-MS) to calculate ILCR. These data served as inputs for the ANN models. The conventional ANN model yielded a coefficient of determination (R²) of 0.73 and a mean squared error (MSE) of 0.0100. In contrast, the proposed method achieved significantly improved performance, with an R² of 0.93 and an MSE of 0.0031.

This improvement highlights the efficacy of the novel preprocessing technique, which optimally integrates meteorological parameters, particularly wind-related factors, into the modelling framework. Moreover, the proposed model’s ability to generalize across multiple sites allows it to be trained on larger datasets, thereby enhancing its robustness and reliability for predicting ILCR in various urban areas.

The study's findings emphasize the importance of refining input parameter representation in ANN-based environmental risk models to achieve superior accuracy and broader applicability. This work not only demonstrates the feasibility of using advanced AI techniques to assess public health risks but also offers a scalable solution for multi-site applications, paving the way for better-informed environmental and public health policies.