FeatTS-Detect : Features-Driven Time Series Anomaly Detection

With the rapid growth of sensors in Cyber-Physical Systems such as clinical data, industrial systems and data centres, there is an increasing need to monitor these devices to secure them against anomalies. This is particularly the case for streaming clinical data. Indeed, the timeliness revelation of anomalies in these data can save the patient’s life.
Time series anomaly detection has been a perennially important topic in data science, but in recent years there has been an explosion of interest in this topic.
The main objective of the project is to implement a system that permits to discover anomaly points and the anomaly subsequence of points in the time series data through a system that evaluates an anomaly considering a set of features extracted among the time series. The features have obtained good performance for clustering of time series. Therefore, our idea is to extend this technique for detecting anomalies.

Faculty Supervisor:

Raymond Ng

Student:

Partner:

Université Claude Bernard Lyon 1

Discipline:

Computer science

Sector:

Education

University:

The University of British Columbia

Program: