BIUSTRE

A multi layer perceptron classifier for radio emission sources associated with galaxy clusters

Show simple item record

dc.contributor.supervisor Hlomani, Hlomani
dc.contributor.supervisor Chibueze, James Okwe
dc.contributor.author Dichaba, Keletso Bontle
dc.date.accessioned 2025-08-25T10:16:05Z
dc.date.available 2025-08-25T10:16:05Z
dc.date.issued 2023-01
dc.identifier.citation Dichaba, K.B. (2023) A multi layer perceptron classifier for radio emission sources associated with galaxy clusters, Masters Theses, Botswana International University of Science and Technology: Palapye en_US
dc.identifier.uri https://repository.biust.ac.bw/handle/123456789/641
dc.description Thesis (MSc Computer Science)--Botswana International University of Science and Technology, 2023 en_US
dc.description.abstract The Square Kilometre Array (SKA) is an emerging next-generation radio telescope project estimated to be completed a decade from now. Once completed it will make it easier and faster to survey space in greater detail. This will produce petabytes of data which will be higher by a factor of one hundred when compared to existing ra dio telescopes presenting an array of scientific and technical challenges. To address the challenges individual and combined solutions are required by utilizing it’s precur sors. Processing data in real time has since become an integral part of future data processing pipelines that will be designed and implemented with the aim of identify ing and selecting important data for further analysis from the bulk of the incoming data while discarding what is not needed or considered unnecessary. It also makes the traditional manual handling of data an unfeasible approach. To this end this study explores the automatic classification of radio emission sources associated with galaxy clusters through the use of a Multi Layer Perceptron (MLP) as a way to eliminate the manual inspection that is often required to confirm individual sources and determine its feasibility as a preliminary eliminator of unnecessary data. This was achieved by implementing an MLP to train on over 400 000 radio emission sources and then predicting the class label on over 100 000 sources. The three class labels were ‘S’ which denoted a single Gaussian source in a given island, ‘C’ which denoted a single Gaussian source in an island with another source and ‘M’ which denoted a multi Gaussian radio source in a given island. The MLP achieved an average accuracy of 0.89, a precision of 0.82, a recall of 0.81 and a F1_score of 0.80. The results proved that an MLP serves as a good classifier for radio emission sources associated with radio galaxies. en_US
dc.description.sponsorship Botswana International University of Science and Technology (BIUST) en_US
dc.language.iso en en_US
dc.publisher Botswana International University of Science and Technology (BIUST) en_US
dc.subject Square Kilometre Array (SKA) en_US
dc.subject Radio emission sources en_US
dc.subject Multi Layer Perceptron (MLP) en_US
dc.subject Automatic classification en_US
dc.subject Real-time processing en_US
dc.subject Galaxy clusters en_US
dc.subject Machine learning en_US
dc.subject Gaussian source en_US
dc.subject F1_score en_US
dc.subject Data processing pipelines en_US
dc.title A multi layer perceptron classifier for radio emission sources associated with galaxy clusters en_US
dc.description.level msc en_US
dc.description.accessibility unrestricted en_US
dc.description.department cis en_US


Files in this item

Icon
Name: Dichaba MSc_Thesi ...
Size: 5.558Mb
Format: PDF
View/Open

This item appears in the following Collection(s)

  • Faculty of Sciences
    This collection is made up of electronic theses and dissertations produced by post graduate students from Faculty of Sciences

Show simple item record

Search BIUSTRE


Browse

My Account