Student Number : 0213053E
MSc research report -
School of Electrical and Information Engineering -
Faculty of Engineering and the Built Environment / One of the risks that have a great impact on society is military con-
°ict. Militarised Interstate Dispute (MID) is de¯ned as an outcome of
interstate interactions which result in either peace or con°ict. E®ective
prediction of the possibility of con°ict between states is a good decision
support tool. Neural networks (NNs) have been implemented to predict
militarised interstate disputes before Marwala and Lagazio [2004]. Sup-
port Vector Machines (SVMs) have proven to be very good prediction
techniques in many other real world problems Chen and Odobez [2002];
Pires and Marwala [2004]. In this research we introduce SVMs to predict
MID. The results found show that SVM is better in predicting con°ict
cases (true positives) without e®ectively reducing the number of correctly
classi¯ed peace (true negatives) than NN. A sensitivity analysis for the
in°uence of the dyadic (explanatory) variables shows that NN gives more
consistent and easy to interpret results than SVM. Further investigation
is required with regards to the sensitivity analysis of SVM.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:wits/oai:wiredspace.wits.ac.za:10539/1515 |
| Date | 31 October 2006 |
| Creators | Habtemariam, Eyasu A. |
| Source Sets | South African National ETD Portal |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | 320707 bytes, application/pdf, application/pdf |
Page generated in 0.0018 seconds