Explainable sequence-to-sequence GRU neural network for pollution forecasting
Abstract The goal of pollution forecasting models is to allow the prediction and control of the air quality. Non-linear data-driven approaches based on deep neural networks have been increasingly used in such contexts showing significant improvements w.r.t. more conventional approaches like regressi...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-06-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-023-35963-2 |
| _version_ | 1797795805007970304 |
|---|---|
| author | Sara Mirzavand Borujeni Leila Arras Vignesh Srinivasan Wojciech Samek |
| author_facet | Sara Mirzavand Borujeni Leila Arras Vignesh Srinivasan Wojciech Samek |
| author_sort | Sara Mirzavand Borujeni |
| collection | DOAJ |
| description | Abstract The goal of pollution forecasting models is to allow the prediction and control of the air quality. Non-linear data-driven approaches based on deep neural networks have been increasingly used in such contexts showing significant improvements w.r.t. more conventional approaches like regression models and mechanistic approaches. While such deep learning models were deemed for a long time as black boxes, recent advances in eXplainable AI (XAI) allow to look through the model’s decision-making process, providing insights into decisive input features responsible for the model’s prediction. One XAI technique to explain the predictions of neural networks which was proven useful in various domains is Layer-wise Relevance Propagation (LRP). In this work, we extend the LRP technique to a sequence-to-sequence neural network model with GRU layers. The explanation heatmaps provided by LRP allow us to identify important meteorological and temporal features responsible for the accumulation of four major pollutants in the air ( $$\text {PM}_{10}$$ PM 10 , $$\text {NO}_{2}$$ NO 2 , $$\text {NO}$$ NO , $$\text {O}_{3}$$ O 3 ), and our findings can be backed up with prior knowledge in environmental and pollution research. This illustrates the appropriateness of XAI for understanding pollution forecastings and opens up new avenues for controlling and mitigating the pollutants’ load in the air. |
| first_indexed | 2024-03-13T03:23:34Z |
| format | Article |
| id | doaj.art-c7fb1ebf47dd481baab897b74073da98 |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-03-13T03:23:34Z |
| publishDate | 2023-06-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-c7fb1ebf47dd481baab897b74073da982023-06-25T11:13:42ZengNature PortfolioScientific Reports2045-23222023-06-0113111810.1038/s41598-023-35963-2Explainable sequence-to-sequence GRU neural network for pollution forecastingSara Mirzavand Borujeni0Leila Arras1Vignesh Srinivasan2Wojciech Samek3Department of Artificial Intelligence, Fraunhofer Heinrich Hertz InstituteDepartment of Artificial Intelligence, Fraunhofer Heinrich Hertz InstituteDepartment of Artificial Intelligence, Fraunhofer Heinrich Hertz InstituteDepartment of Artificial Intelligence, Fraunhofer Heinrich Hertz InstituteAbstract The goal of pollution forecasting models is to allow the prediction and control of the air quality. Non-linear data-driven approaches based on deep neural networks have been increasingly used in such contexts showing significant improvements w.r.t. more conventional approaches like regression models and mechanistic approaches. While such deep learning models were deemed for a long time as black boxes, recent advances in eXplainable AI (XAI) allow to look through the model’s decision-making process, providing insights into decisive input features responsible for the model’s prediction. One XAI technique to explain the predictions of neural networks which was proven useful in various domains is Layer-wise Relevance Propagation (LRP). In this work, we extend the LRP technique to a sequence-to-sequence neural network model with GRU layers. The explanation heatmaps provided by LRP allow us to identify important meteorological and temporal features responsible for the accumulation of four major pollutants in the air ( $$\text {PM}_{10}$$ PM 10 , $$\text {NO}_{2}$$ NO 2 , $$\text {NO}$$ NO , $$\text {O}_{3}$$ O 3 ), and our findings can be backed up with prior knowledge in environmental and pollution research. This illustrates the appropriateness of XAI for understanding pollution forecastings and opens up new avenues for controlling and mitigating the pollutants’ load in the air.https://doi.org/10.1038/s41598-023-35963-2 |
| spellingShingle | Sara Mirzavand Borujeni Leila Arras Vignesh Srinivasan Wojciech Samek Explainable sequence-to-sequence GRU neural network for pollution forecasting Scientific Reports |
| title | Explainable sequence-to-sequence GRU neural network for pollution forecasting |
| title_full | Explainable sequence-to-sequence GRU neural network for pollution forecasting |
| title_fullStr | Explainable sequence-to-sequence GRU neural network for pollution forecasting |
| title_full_unstemmed | Explainable sequence-to-sequence GRU neural network for pollution forecasting |
| title_short | Explainable sequence-to-sequence GRU neural network for pollution forecasting |
| title_sort | explainable sequence to sequence gru neural network for pollution forecasting |
| url | https://doi.org/10.1038/s41598-023-35963-2 |
| work_keys_str_mv | AT saramirzavandborujeni explainablesequencetosequencegruneuralnetworkforpollutionforecasting AT leilaarras explainablesequencetosequencegruneuralnetworkforpollutionforecasting AT vigneshsrinivasan explainablesequencetosequencegruneuralnetworkforpollutionforecasting AT wojciechsamek explainablesequencetosequencegruneuralnetworkforpollutionforecasting |