Khanh Luong

    Project details

    Title: Clustering methods on Multi-type Relational Data

    With the rapid growth of computational technology, simultaneously clustering many interrelated data types into groups has become a promising research area. Several clustering methods on two-type or multi-type relational data have been proposed that rely on inter-type relationships or intra-type relationships or both show effectiveness compared to traditional ones.

    This research project will explore, empirically analyse and develop clustering methods on multi-type relational data focusing on relationships, constraints and methods to overcome shortcomings of existing methods. This research thesis will provide a new perspective on analysis of this emerging and significant type of data.

    Studying clustering methods especially clusteirng methods on multi-type relational data. Multi-type relational data is a new emmering data that can embed rich information as it includes many types of objects and can have many kinds of relationships. Clustering on multi-type relational data has been proved to find more meaningful clusters as compared to clustering on “flat” data that can only capture data and their features. Yet due to the requirement to take all possible intrinsic relationships that contain real and much valuable information into account and

    the natural complexity of heterogeneous data, clustering on MTRD is a challenging problem and worthwhile to be paid meticulous attention.

    In my PhD thesis, clustering methods focus on combining Nonnegative Matrix Factorization framework and manifold learning to find meaningful low-dimensional space of original data and looking for clusters in the new representations.