It is right next to the air quality forecast graph. The percentage factor is visible in the side panel after selecting the measuring station on the map. The accuracy value is calculated for each Airly station, based on measurement data and forecasts collected from the last 14 days. We monitor the accuracy of our forecast on an ongoing basis and provide this information in our applications.
In our opinion, this proves the high effectiveness of our predictive model. However, the difference between the expected and actual air quality is rarely big enough to make the forecast completely inaccurate. For example, we may predict very good air, but in reality, it will be "just" good. Similarly, when our forecast predicts a very high level of pollution, almost certainly smog awaits us - put on your masks and preferably do not leave the house! In the remaining 5% of cases, our model may be wrong, but usually, it is off by only 1 CAQI level. In other words, when our model predicts very good air, there is a 95% chance it will be so. This means that when forecasting air quality, in 95 cases out of 100, the absolute error of our prediction (relative to later-measured values) is not higher than 25 on the CAQI scale, i.e. According to our calculations, the accuracy of our air quality forecast is over 95%.
0 kommentar(er)
