We are happy to announce that the LV Datathon is taking place on 20 and 22 April 2022 in person at the University of Bristol for Compass and IAI CDT students.

Students have been provided with the information below:

European and Spanish legislation set hourly limits for certain pollutants, that are enforced with traffic restrictions. This task will look into forecasting pollution levels in Madrid city centre by using pollutant readings from the monitoring station closest to the city centre located at the latitude / longitude coordinate [40.416775, -3.703790], data recordings will be provided for all the monitoring stations in the city.

Forecasts should be provided in mg per cubic metre for 3 key pollutants: carbon monoxide, sulphur dioxide and nitrogen dioxide; two weeks in advance. Three years’ worth of readings will be provided at the beginning of the exercise to build the forecasting models.

The goal of the Datathon is both to gain predictive power for such data, as well as to be able to understand such complex multivariate data based on three criteria:

1.Predictive accuracy: How accurately can students forecast the two-week pollutant curve for the closest station? The forecasting performance will be assessed using RMSE and WMAPE metrics

2.Correlation structure: Can students understand the pattern between curves across stations? Does predictive performance for one station improve with access to others, and are mistakes correlated?

3.Visualisation: Can students identify novel visualisation approaches that add insight?

Each team will be awarded a final score for each of the following criteria:

1.Predictive performance

2.Practicality / usability

3.Explainability / interpretability

4.Innovation