Context:

One of our clients wanted to build a predictive system that uses biometric data and self-reported questionnaires to forecast migraine onset. The system integrates deep learning models in Python and a real-time dashboard in R Shiny to visualize predictions and collect feedback.

Resolution:

To automate the process, our team stored Data in AWS S3 buckets that were Organized by participant ID, timestamp, and signal type. The Data Preprocessing was done in Python that signalled smoothing and noise reduction, created Time-series alignment across variables. Deep learning model used multi-input neural network, LSTM layers for time-series signals and Dense layers for static features.

Dashboard prototype using R Shiny displayed real-time prediction, tracked biometric trends, and collected feedback.

Result:

The system added anomaly detection for rare migraine types, enabled federated learning for privacy-preserving updates. It showed HRV dips and elevated cortisol levels often precede migraines, also questionnaire data improved model sensitivity, and the feedback loop enhanced personalization and accuracy.