To monitor for panic attacks via heart rate spikes, continuous heart rate tracking from your watch, linked to your phone, is essential. We utilize WCSessions, an iOS library enabling the connection between your Apple Watch and phone. In Xcode, using Swift, we facilitate data transfer to our server's Firebase Realtime Database, where the heart rate information is stored for analysis.

After collecting the heart rate data, we process it through our machine learning model, a combination of Random Forest and Time Series (LSTM) models, coded in Python and deployed on Firebase. The LSTM model analyzes heart rate fluctuations over time, while our Random Forest model classifies these fluctuations as either indicative of a panic attack or not.

If our machine learning model detects a panic attack, it sends a binary value of 0 (no panic attack) or 1 (panic attack detected). On detecting a panic attack, a notification is sent to your phone, guiding you to our user interface. The interface first confirms the panic attack, then leads you through five grounding exercises and five breathing exercises. It offers the option to repeat the exercises or conclude the session.

When you confirm a panic attack through the app, it triggers a message to be sent. Initially, upon logging into the app, you have the option to enter a guardian or parent's phone number, which is then stored in our Firebase Realtime Database. Upon confirmation of a panic attack, our Swift code initiates Firebase functions. These functions, written in Java and integrated with a third-party SMS messaging service called Vonage, retrieve the stored phone number from the database and send a message to your specified contact.