Masters Degrees (Computer Science)

Permanent URI for this collection

https://scholar.sun.ac.za/handle/10019.1/96339

Browse

Browsing Masters Degrees (Computer Science) by Subject "Architecture and Data Completion"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Scaling the ConceptCloud browser to very large semi-structured data sets: architecture and data completion
    (Stellenbosch : Stellenbosch University, 2020-12) Berndt, Joshua; Fischer, Bernd; Britz, K.; Stellenbosch University. Faculty of Science. Dept. of Mathematical Sciences. Division Computer Science.
    ENGLISH ABSTRACT: Semi-structured data sets such as product reviews or event log data are simultaneously becoming more widely used and ever larger. This thesis describes ConceptCloud, a exible, interactive browser for semi-structured datasets, with a focus on the improvements made to accommodate larger datasets, more intuitive data representation and the enrichment of the underlying data by way of data-imputation. ConceptCloud makes use of an intuitive tag cloud visualisation viewer in combination with an underlying concept lattice to provide a formal structure for navigation through datasets without prior knowledge of the structure of the data or compromising scalability. This scalability is achieved by the implementation of architectural changes to increase the system's resource efficiency. These changes are demonstrated by way of a case study on a dataset of wine reviews. Semi-structured data sets such as product reviews or event log data often contain a geolocation aspect: for example, the location of the winery for wine reviews, or the accident location for traffic data. In this thesis, I describe ConceptCloud extensions which allow for the rendering of specialised geolocation data while providing alternate navigation paths through the dataset. I show that using biclusters can make the navigation bidirectional, and demonstrate this approach on a crime data set making use of a geolocation specialised map viewer. Semi-structured data often contains implicit information which will be useful in driving data exploration if made explicit. I take advantage of domain ontologies to both allow implicit data in each input data set to be made explicit and verify and correct inconsistencies allowing for better data exploration. I demonstrate this approach with a continuation of the wine case study.