|
Activity Number:
|
331
- Cluster Detection in Big Data
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Mental Health Statistics Section
|
|
Abstract #313186
|
|
|
Title:
|
Unsupervised Filtering Algorithm for Passively Collected Smartphone Data
|
|
Author(s):
|
Hongzhe Zhang* and Samprit Banerjee and Jihui Lee
|
|
Companies:
|
Weill Cornell Medicine and Weill Medical College, Cornell University and Weill Cornell Medicine
|
|
Keywords:
|
smartphone;
unsupervised;
filtering;
mHealth;
depression;
wearable
|
|
Abstract:
|
Recent advances in mobile technology has expanded a market for digital health, and smartphones has gained its popularity as a source for a variety of data including daily step count and conversation or sleep duration. These measures from smartphones play an important role in efficiently monitoring the user’s activity and social interaction. A key issue with passively collected mHealth data is that their quality heavily depends on the users’ interaction with their device. We developed a novel two-step unsupervised filtering algorithm to identify the data points when users have low smartphone engagement and therefore unreliable measurements. The first step is to assign weak labels by jointly considering measures from device sensors (e.g. accelerometer, GPS, and microphone) via principal component analysis (PCA). In the second step, we implement an adjusted k-nearest neighbors method to re-calibrate the weak labels.
|
Authors who are presenting talks have a * after their name.
