Dynamic Time Warping Based Clustering Anonymization Approach for Patient Time Series

Overview

In this digital age, we give away our data everyday; by accepting website cookies or wearing smartwatches that track everything from heart rate to sleep cycles. The data we generate can be used to design solutions to problems faced across all fields of science but cannot be released without the guarantee of anonymity for subjects involved. Trying to strike a balance between privacy and the meaningfulness of the data, especially with time series data (data collected over a period), has stood as a challenge for researchers. This project aims to evaluate the efficacy of a novel approach: using clustering to group similar subjects and generate an averaged set of values (centroid), that in theory should maintain anonymity. The clustering method was evaluated by training a forecasting model to predict future values based on the centroid and compare with the actual values.

Awards (2)

• Bronze Medal
• Selected for CWSF 2026

Competition history

• CWSF 2026 Health & Wellness

Related projects

Closest projects by meaning, across every fair and year in the corpus.