Exploratory analysis of time series data: Detection of partial similarities, clustering, and visualization

• We propose a new method for analyzing time series data.
• The method detects partial similar patterns in simultaneously occurring time series data at different locations.
• A graphical representation helps us understand the similarity between the data and classify them into smaller subgroups.
• Numerical measures evaluate the effectiveness of clusters and provide a means of testing their statistical significance.

A new exploratory method for analyzing time series data is proposed. A computational algorithm detects partial similarities between simultaneously occurring time series data and clusters the data into groups based on their similarities. A graphical representation that visualizes the data clustering process helps us understand similarity between time series data and classifies them into smaller subgroups. Numerical measures evaluate the effectiveness of clusters and provide a means for testing their statistical significance. The proposed method was applied to an analysis of the change of population distribution during a day in Salt Lake County, Utah, USA. This application supports the technical soundness of the method and provides empirical findings.