Barometric Crash Detector

[AreTaj]

Objective

Implement a feature to detect rapid drops in barometric pressure, which are known migraine triggers (hypoxia/expansion mechanism).

Technical Implementation

• Source: Hourly weather data (Open-Meteo).
• Metric: max_3h_pressure_drop
• Formula: Calculate the 3-hour rolling difference for every hour in a 24h window, then take the minimum (most negative) value.
• Why: Daily averages hide these short-term "crash" events.

Acceptance Criteria

• data_loader.py calculates this feature from hourly arrays.
• Feature is correctly fed into the Gradient Boosting model.
• Unit test verifies that a sharp drop (e.g. 1020 -> 1000 in 3h) registers correctly.