Online Program

Phases in Dynamic Systems: Cluster Analysis of Time Series Data (302724)

Keywords: time series, cluster analysis, segmentation, seasonality

The use of cluster analysis to identify successive phases in a dynamic process is illustrated. While static characteristics are often used to identify clusters or population segments, evolving systems may be better described by how things change over time. A medical patient may be classified by the amount of time since an important event such as a diagnosis, economic activity may be segmented by stages in an economic cycle, and neighborhoods may be grouped by stages in generational evolution. An important success of this technique is in astrostatistics, where a supernova is classified by how the amount of light it produces changes over time. Examples are given in biostatistics, econometrics, meteorology, and astrostatistics.