Detection of suspicious and/or abnormal real-time events from textual live data feeds

Social media and other real-time messaging applications represent valuable sources of real-time information that remain untapped by many service operators. The project is aimed at developing methodology for detecting suspicious and/or abnormal real-time events from textual live data feeds, based on predictive and/or anomaly detection algorithms applied to time series and text features.
TRT Canada is therefore interested in developing algorithms that will be able to recognize such events based on similarities with past events in order to address the mentioned scenario.
The research consists of evaluating some of the state-of-the-art methods in textual analytics in text stream processing and neural network techniques, such as deep learning and convolutional neural network. The experiments are designed to measure effectiveness of these methods at detecting anomalies and new events, event classification, and identification of event features, such as actors, locations, and time characteristics.

Faculty Supervisor:

Stan Matwin

Student:

Gashin Ghazizadeh

Partner:

Discipline:

Computer science

Sector:

Information and communications technologies

University:

Program:

Accelerate