Proper Orthogonal Decomposition Methods for the Analysis of Real-Time Data: Exploring Peak Clustering in a Secondhand...

Area of Science:

• Environmental Science
• Data Science
• Public Health

Background:

• Secondhand smoke exposure is a significant public health concern.
• Monitoring airborne particles in households can help assess exposure levels.
• Classifying complex time-series data from monitors presents analytical challenges.

Purpose of the Study:

• To develop and evaluate a method for classifying peaks in time-series data from household air particle monitors.
• To assess the effectiveness of a clinical intervention aimed at reducing secondhand smoke exposure.
• To correlate identified particle peak patterns with intervention outcomes.

Main Methods:

• Utilized proper orthogonal decomposition (POD) for dimensionality reduction of time-series data.
• Applied a k-means clustering algorithm to categorize data peaks into distinct clusters.
• Estimated parameters for a physics-based model of airborne particles for each identified cluster.
• Assessed classification accuracy of the POD/clustering method on generated peaks.

Main Results:

• The POD/k-means method successfully classified data peaks into two distinct clusters.
• Following an intervention with aversive feedback, the proportion of short-duration, attenuated peaks increased from 38.8% to 96.6%.
• The POD/clustering approach achieved >60% accuracy in identifying peaks generated from physics-based model distributions.

Conclusions:

• The developed POD/k-means method is effective for classifying airborne particle peaks in household monitoring data.
• The clinical intervention significantly altered particle peak characteristics, indicating reduced exposure.
• This approach provides a robust tool for analyzing environmental exposure data and evaluating public health interventions.