The Best Glossary of Air Quality

To measure these pollutants precisely, Libelium Air Quality Station uses Machine Learning. Libelium engineers have used statistical and Artificial Intelligence techniques. Machine Learning algorithms were trained with loads of data to obtain advanced models that accurately enhance sensor quality. ML models ‘learn’ as a result of the observation of tons of data and spontaneously find a solution. Then, the model makes predictions, finds patterns, and draws conclusions without human interaction.

Libelium engineers have used statistical and Artificial Intelligence techniques. Machine Learning algorithms were trained with loads of data to obtain advanced models that accurately enhance sensor quality

The goal is to get the best R² in the market. The term R² is a very common and vital concept, but it can be complex and misleading. The coefficient of determination (R²) measures a comparison between an air quality device and another taken as a reference. The closer to 1, the better the device imitates the performance of the reference device. 

R² is always an important indicator of the quality of the sensor. Remember that, depending on the equation applied, R² does not include the error, and it must be read with other indicators, for example with error parameters like the MAE (Mean Absolute Error), RMSE (Root-Mean-Square Error) or the statistical distribution of the error.  The RMSE allows us to evaluate the dispersion of the errors too because it penalizes especially the longest distances. This metric is always highly valuable since it describes well the accuracy of a sensor.

Unlike other data providers, the R² parameter calculated by Libelium is directly related to the RMSE.