Emergency alerting system and method

In example embodiments, a machine, including one or more processors and a memory, tracks, by communicating over a network with a plurality of devices associated with a user, activity of the user. The machine develops, using the one or more processors, an activity model for the user based on the tracked activity of the user. The machine determines, an anomaly in a current activity of the user relative to the developed activity model, the anomaly having a type and a duration. The machine calculates, based on the type and the duration of the anomaly, a confidence value corresponding to whether the user needs assistance and a severity value indicating severity of the user's need for assistance. The machine provides, to an emergency contact and via the network, an alert indicating that the user needs assistance based on the confidence value or the severity value.