AirNest

The thin bar at the bottom of each card represents the danger probability — a value between 0% and 100% that reflects how confident the model is that air quality conditions will reach a warning threshold on that day. A longer, red bar indicates higher predicted risk; a short green bar indicates the day is expected to remain within safe limits. The percentage label next to the bar shows the exact confidence value.

Two models work together to produce each daily forecast. Prophet (developed by Meta) is a time-series forecasting model used to predict temperature and humidity for the coming week. It learns seasonal patterns and trends from historical data collected by the station and cross-references them with weather forecasts provided by Open-Meteo. A Random Forest classifier then takes those predicted environmental conditions, alongside recent MQ sensor readings, and determines whether each day is likely to be SAFE or DANGER. Random Forest works by building a large number of decision trees on different subsets of the training data and combining their votes into a final prediction, which makes it robust to noise and outliers in the sensor data.