Graph-based complex pattern mining in electronic health records
Project Leader: Pauline Lin
Collaborators: Prof Uwe Aickelin
Primary Contact: Pauline Lin (email@example.com)
Keywords: cancer; data mining; data structures; health and bioinformatics; machine learning
Disciplines: Computing and Information Systems
Electronic health records are information rich, contain complex contextual relationships among observations, with rich annotations and a semi-time-series nature. Such data can be modelled as dynamic high-dimensional networks. Inter-dependent patterns in dynamic networks are interesting and often not obvious. Many known patterns exist in network structures (such as power laws in degree distributions, eigenvalues of the agency matric, and triangle laws). Current pattern mining methods include analysing graph specific patterns (eg, page rank, random walk and belief propagation) and learning of various network representations for mining tasks. This is a project which focuses on exploring new network-based methods for pattern mining on real electronic health records where the data are rich and yet noisy.