Discover anomaly signatures from time series data of telecommunication networks

Failures in a telecommunication network harm the communication quality. Once happened, if the system cannot solve it by self-healing, such anomaly may even result in serious problem and result in massive economic loss. In this project, we will design and develop a system to predict these failures in advance using the status values of the hardware facilities. Our goal is to build a completed data processing, model building and training system to predict facility failures automatically for production-level deployment with strict evaluation criteria (precision > 80%, which means for all the positive prediction our model gives, at least 80% of them is correct; recall > 10%, which means for all real failures in the network, our model could predict at least 10% of them). The success of this project will expand Ciena’s capability to develop superior products for anomaly prediction services.

Faculty Supervisor:

Yan Liu

Student:

Wenjie Du

Partner:

Ciena Canada

Discipline:

Engineering - computer / electrical

Sector:

Information and cultural industries

University:

Concordia University