Includes bibliographical references. / Cluster analysis is becoming an increasingly popular method in modern finance because of its ability to summarise large amounts of data and so help individual and institutional investors to make timeous and informed investment decisions. This is no less true for investors in smaller, emerging markets - such as the Johannesburg Stock Exchange - than it is for those in the larger global markets. This study examines the application of two clustering techniques to the Johannesburg Stock Exchange. First, the application of Salvador and Chan's (2003) L method stopping rule to a hierarchical clustering of time series return data was analysed as a method for determining the number of latent groups in the data set. Using Ward's method and the Euclidean distance function, this method appears to be able detect the correct number of clusters on the JSE. Second, the ability of three different clustering algorithms to generate consistent clusters and cluster members over time on the Johannesburg Stock Exchange was analysed. The variation of information was used to measure the consistency of cluster members through time. Hierarchical clustering using Ward's method and the Euclidean distance measure proved to produce the most consistent results, while the K-means algorithms generated the least consistent cluster members.
Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:uct/oai:localhost:11427/8562 |
Date | January 2014 |
Creators | Tully, Robyn |
Publisher | University of Cape Town, Faculty of Commerce, Department of Finance and Tax |
Source Sets | South African National ETD Portal |
Language | English |
Detected Language | English |
Type | Master Thesis, Masters, MPhil |
Format | application/pdf |
Page generated in 0.0101 seconds