Year: 2016
Type: dissertation
Abstract: Big data is a growing area of research with some important research challenges that motivate \nour work. We focus on one such challenge, the variety aspect. First, we introduce \nour problem by defining heterogeneous data as data about objects that are described by \ndifferent data types, e.g., structured data, text, time-series, images, etc. Through our work \nwe use five datasets for experimentation: a real dataset of prostate cancer data and four \nsynthetic dataset that we have created and made them publicly available. Each dataset \ncovers different combinations of data types that are used to describe objects. Our strategy \nfor clustering is based on fusion approaches. We compare intermediate and late fusion \nschemes. We propose an intermediary fusion approach, Similarity Matrix Fusion (SMF), \nwhere the integration process takes place at the level of calculating similarities. SMF produces \na single distance fusion matrix and two uncertainty expression matrices. We then \npropose a clustering algorithm, Hk-medoids, a modified version of the standard k-medoids \nalgorithm that utilises uncertainty calculations to improve on the clustering performance. \nWe evaluate our results by comparing them to clustering produced using individual elements \nand show that the fusion approach produces equal or significantly better results. \nAlso, we show that there are advantages in utilising the uncertainty information as Hkmedoids \ndoes. In addition, from a theoretical point of view, our proposed Hk-medoids \nalgorithm has less computation complexity than the popular PAM implementation of the \nk-medoids algorithm. Then, we employed late fusion that aggregates the results of clustering \nby individual elements by combining cluster labels using an object co-occurrence \nmatrix technique. The final cluster is then derived by a hierarchical clustering algorithm. \nWe show that intermediate fusion for clustering of heterogeneous data is a feasible and \nefficient approach using our proposed Hk-medoids algorithm.
Author: Aalaa Mojahed
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: King Abdulaziz University
OpenAlex ID: W2529664025