Anomaly Detection for Dynamic Graphs
Adrian Trujillo, Austin Thao, Spencer Ortega
We will work with our mentor, Dr. Kumar Sricharan, to create a web application that will detect significant changes in time-evolving graphs. Our application will be designed for data analyst with minimal programming experience and will visualize the anomalous changes in user provided dynamic data.
We want to approach the project from an Agile Software Development Life Cycle to get constant feedback from our mentor. Leveraging existing research as a resource, we will find more temporal graph data and synthesize relational data sets.Then, we plan to use Python to create a module to ingest that graph data. Once we have the ingestion module done, we will utilize the CAD (Commute-time based Anomaly detection in Dynamic graphs) algorithm to detect anomalous node relationships. Using the Euclidean graph model, we will visualize the detected anomalies in a web based application.
We will create a web application that will take in user temporal graph data and detect the significant changes in node relationships over time. Once the anomalies are found, the user will be able to view all the anomalous node relations in a user-friendly interface.