Periodic Pattern Detection
An important factor in understanding human behavior is understanding patterns of events over time. The members of a department generate important sales meeting notes weekly; two friends usually discuss tennis or rock climbing every Friday afternoon; an account manager talks to an important client once a week.
These events show the heartbeat of an organization, and help show what its priorities and processes are. Variations from these patterns can show important changes in the environment—a change in business strategy, or an employee becoming disgruntled and disengaged.
How do you find such patterns against the background data, without prior knowledge of the actors and topics involved? Cataphora’s periodic pattern detection technology is up to the task. This requires uncovering quasi-periodic patterns from a deluge of streaming and constantly updated data. Essentially, this is the unsupervised identification of “relevant” classes of electronic events that share similar semantics and display structural or temporal regularities.
The units of data to analyze, which we call electronic events, must go beyond the simple one-to-one correspondence to electronic documents such as emails or spreadsheets. A seemingly innocuous mention of a sensitive security topic on unauthorized channels, or a wide set of communications and documents emitted during a major corporate activity (hiring process, SEC filing, etc.) are both valid examples of electronic events that transcend the scope of a single document.
Automating the selection of the relevant units, features and semantics to study is therefore one of the central challenges.
Read about other Cataphora technologies:
