Data and code associated with the publication: Relapse prediction using wearable data through convolutional autoencoders and clustering for patients with psychotic disorders

Our work focused on developing a personalized approach using neural-network-based anomaly detection and clustering to predict relapse for patients with psychotic disorders. We used a dataset provided by e-Prevention grand challenge (https://robotics.ntua.gr/eprevention-sp-challenge/), containing physiological signals for 10 patients monitored over 2.5 years. We created 2-dimensional multivariate time-series profiles containing activity and heart rate variability metrics, extracted latent features via convolutional autoencoders, and identified relapse clusters. In this repository, we shared our aggregated 2D multivariate time-series profiles for users to understand the structure of processed data and make it easier to run our demo code. (2025-03)