Abstract
BackgroundAdvances in measurement technology are producing increasingly time-resolved environmental exposure data. We aim to gain new insights into exposures and their potential health impacts by moving beyond simple summary statistics (e.g., means, maximums) to characterize more detailed features of high-frequency time-series data. MethodsThis study proposes a novel variant of the Self-Organizing Map (SOM) algorithm called Dynamic Time Warping Self-Organizing Map (DTW-SOM) for unsupervised pattern discovery in time series. This algorithm uses DTW, a similarity measure for sequential data that optimally aligns interior patterns, both as the similarity measure and for training the neural network.ResultsWe applied DTW-SOM to a panel study monitoring indoor and outdoor residential environmental exposures for 10 patients with asthma from 7 households near Salt Lake City, Utah; each patient was followed for up to 373 days. Compared to other SOM algorithms using Euclidean distance, the DTW-SOM algorithm maintained the topological properties of the input time series and generated more detailed diurnal patterns. We observed seasonal patterns in outdoor temperature and distinct patterns of indoor peak PM2.5 exposure, which was likely linked to both combustion sources and days with increased inhaler usage. ConclusionsThe new algorithm, DTW-SOM, better preserved the topology relationship of time-series data and better summarized time-series patterns as compared to the original version of SOM.