This invention presents a method for efficiently detecting abnormal changes in the underlying statistics of high-dimensional data streams in a completely unsupervised manner.

About

A Method for Unsupervised Change Detection Using Deep Learning

Tech ID: UA19-209

Invention:
This invention is a method of auto-detecting a change in data trends. As a data stream changes in characteristics, it may trigger other events, predict future performance, or assess a business’ health, among other things. Specifically, the invention looks to analyze high dimensional time-series data. This data appears in numerous applications, including healthcare, financial transactions, network traffic measurements, and Internet of Things (IoT). This invention presents a method for efficiently detecting abnormal changes in the underlying statistics of high-dimensional data streams in a completely unsupervised manner.

Background:
Detecting changes in a data stream is an important area of research with many applications. In many applications, data is not static but arrives in data streams. Besides the algorithmic difference between processing data streams and static data, there is another significant difference. Change detection can be broken down into the following two categories: Value Change Detection and Aggregation Change Detection.

Key Benefits

- Adaptable to many sectors of the economy, including emerging sectors

Applications

- Change detection for healthcare
- Social network change detection
- Change detection for Internet of Things

Purchase a license for full unlimited access to all innovation profiles on LEO

  • Direct connection to thousands of more innovations
  • Access to market Experts and Universities
  • Filter relevant solutions into your own dedicated Network