Skip to Main Content
Data mining is a term encompassing many methods. In this work unsupervised learning, or clustering, was applied to discover new insights from a public access database that lists major disturbances in the power network of the USA over the last 23 years. Results provide evidence that these disturbances can be placed into a few major groups, which can be characterized by region, cause and severity. This analysis also suggests a tendency for disturbances to occur more frequently in the early afternoon and in July. Statistical analysis confirms this conclusion. Such analysis provides a means to automatically characterize complex data, and may lead to fresh insights, and prove useful in planning and upgrade of infrastructure.