Tshilidzi Marwala | University of Johannesburg, South Africa (original) (raw)
Papers by Tshilidzi Marwala
2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence, 2013
Communications in Computer and Information Science, 2012
2009 International Conference on Machine Learning and Applications, 2009
ABSTRACT Accurate classification of caller interactions within Interactive Voice Response systems... more ABSTRACT Accurate classification of caller interactions within Interactive Voice Response systems would assist corporations to determine caller behavior within these telephony applications. This paper details the development of such a classification system for a pay beneficiary application. Fuzzy Inference Systems, Multi-Layer Perceptron, Support Vector Machine and ensemble of classifiers were developed. Accuracy, sensitivity and specificity performance metrics were computed as well as compared for these classification solutions. Ideally, a classifier should have high sensitivity and high specificity. Exceptional results were achieved. The ensemble of classifiers is the preferred solution, yielding an accuracy of 99.17%.
Arxiv preprint arXiv:1110.4296, Jan 1, 2011
South Africa assumes a significant position in the insurance landscape of Africa. The present res... more South Africa assumes a significant position in the insurance landscape of Africa. The present research based upon qualitative and quantitative analysis, shows that it shows the characteristics of a Complex Adaptive Engineering System. In addition, a statistical analysis of risk measures through Value at risk and Conditional tail expectation is carried out to show how an individual insurance company copes under external complexities. The authors believe that an explanation of the coping strategies, and the subsequent managerial implications would enrich our understanding of complexity in business.
Advances in Intelligent Systems and Computing, 2014
ABSTRACT
Proceedings of 2012 IEEE-EMBS International Conference on Biomedical and Health Informatics, 2012
ABSTRACT A lumped transmission line model of the auditory canal is modified to study the influenc... more ABSTRACT A lumped transmission line model of the auditory canal is modified to study the influence of the number of segments used and the variation of the radius on the outer ear frequency response. A new second order polynomial is used to map the radius to the length along the narrow part of the auditory canal. The results found are compared to other literature results and expected theoretical results. The study shows that both an increase and decrease in the number of segments of the ear canal results in a distortion in the frequency response. The results obtained when using the radius-length mapping function show that the response found closely correlates with other literature results. The optimum representation of the frequency response was found when four number of segments and the radius-length mapping function were used. The use of a third order polynomial to further improve the relationship between the radius and the length of the ear canal is suggested for future research work.
IFMBE Proceedings, 2007
ABSTRACT
2004 IEEE International Conference on Systems, Man and Cybernetics (IEEE Cat. No.04CH37583), 2004
ABSTRACT
2009 International Conference on Biomedical and Pharmaceutical Engineering, 2009
... South Africa * Prof. T Marwala has since moved to the University of Johannesburg, South Afric... more ... South Africa * Prof. T Marwala has since moved to the University of Johannesburg, South Africa Abstract - A new artificial Larynx is currently under development at the University of the Witwatersrand, Johannesburg. This device ...
1 in 1982 to describe the first cases of unusual immune system failure that were identified in th... more 1 in 1982 to describe the first cases of unusual immune system failure that were identified in the previous year. The human immunodeficiency virus (HIV) was later identified as the cause of AIDS. Risk factor epidemiology examines the individual demographic and social characteri- stics and attempts to determine factors that place an indi- vidual at risk of acquiring a life-threatening
Journal of Earth System Science, 2014
ABSTRACT The performances of regular support vector machines and random forests are experimentall... more ABSTRACT The performances of regular support vector machines and random forests are experimentally compared for hyperspectral imaging land cover classification. Special characteristics of hyperspectral imaging dataset present diverse processing problems to be resolved under robust mathematical formalisms such as image classification. As a result, pixel purity index algorithm is used to obtain endmember spectral responses from Indiana pine hyperspectral image dataset. The generalized reduced gradient optimization algorithm is thereafter executed on the research data to estimate fractional abundances in the hyperspectral image and thereby obtain the numeric values for land cover classification. The Waikato environment for knowledge analysis (WEKA) data mining framework is selected as a tool to carry out the classification process by using support vector machines and random forests classifiers. Results show that performance of support vector machines is comparable to that of random forests. This study makes a positive contribution to the problem of land cover classification by exploring generalized reduced gradient method, support vector machines, and random forests to improve producer accuracy and overall classification accuracy. The performance comparison of these classifiers is valuable for a decision maker to consider tradeoffs in method accuracy versus method complexity.
2013 BRICS Congress on Computational Intelligence and 11th Brazilian Congress on Computational Intelligence, 2013
Communications in Computer and Information Science, 2012
2009 International Conference on Machine Learning and Applications, 2009
ABSTRACT Accurate classification of caller interactions within Interactive Voice Response systems... more ABSTRACT Accurate classification of caller interactions within Interactive Voice Response systems would assist corporations to determine caller behavior within these telephony applications. This paper details the development of such a classification system for a pay beneficiary application. Fuzzy Inference Systems, Multi-Layer Perceptron, Support Vector Machine and ensemble of classifiers were developed. Accuracy, sensitivity and specificity performance metrics were computed as well as compared for these classification solutions. Ideally, a classifier should have high sensitivity and high specificity. Exceptional results were achieved. The ensemble of classifiers is the preferred solution, yielding an accuracy of 99.17%.
Arxiv preprint arXiv:1110.4296, Jan 1, 2011
South Africa assumes a significant position in the insurance landscape of Africa. The present res... more South Africa assumes a significant position in the insurance landscape of Africa. The present research based upon qualitative and quantitative analysis, shows that it shows the characteristics of a Complex Adaptive Engineering System. In addition, a statistical analysis of risk measures through Value at risk and Conditional tail expectation is carried out to show how an individual insurance company copes under external complexities. The authors believe that an explanation of the coping strategies, and the subsequent managerial implications would enrich our understanding of complexity in business.
Advances in Intelligent Systems and Computing, 2014
ABSTRACT
Proceedings of 2012 IEEE-EMBS International Conference on Biomedical and Health Informatics, 2012
ABSTRACT A lumped transmission line model of the auditory canal is modified to study the influenc... more ABSTRACT A lumped transmission line model of the auditory canal is modified to study the influence of the number of segments used and the variation of the radius on the outer ear frequency response. A new second order polynomial is used to map the radius to the length along the narrow part of the auditory canal. The results found are compared to other literature results and expected theoretical results. The study shows that both an increase and decrease in the number of segments of the ear canal results in a distortion in the frequency response. The results obtained when using the radius-length mapping function show that the response found closely correlates with other literature results. The optimum representation of the frequency response was found when four number of segments and the radius-length mapping function were used. The use of a third order polynomial to further improve the relationship between the radius and the length of the ear canal is suggested for future research work.
IFMBE Proceedings, 2007
ABSTRACT
2004 IEEE International Conference on Systems, Man and Cybernetics (IEEE Cat. No.04CH37583), 2004
ABSTRACT
2009 International Conference on Biomedical and Pharmaceutical Engineering, 2009
... South Africa * Prof. T Marwala has since moved to the University of Johannesburg, South Afric... more ... South Africa * Prof. T Marwala has since moved to the University of Johannesburg, South Africa Abstract - A new artificial Larynx is currently under development at the University of the Witwatersrand, Johannesburg. This device ...
1 in 1982 to describe the first cases of unusual immune system failure that were identified in th... more 1 in 1982 to describe the first cases of unusual immune system failure that were identified in the previous year. The human immunodeficiency virus (HIV) was later identified as the cause of AIDS. Risk factor epidemiology examines the individual demographic and social characteri- stics and attempts to determine factors that place an indi- vidual at risk of acquiring a life-threatening
Journal of Earth System Science, 2014
ABSTRACT The performances of regular support vector machines and random forests are experimentall... more ABSTRACT The performances of regular support vector machines and random forests are experimentally compared for hyperspectral imaging land cover classification. Special characteristics of hyperspectral imaging dataset present diverse processing problems to be resolved under robust mathematical formalisms such as image classification. As a result, pixel purity index algorithm is used to obtain endmember spectral responses from Indiana pine hyperspectral image dataset. The generalized reduced gradient optimization algorithm is thereafter executed on the research data to estimate fractional abundances in the hyperspectral image and thereby obtain the numeric values for land cover classification. The Waikato environment for knowledge analysis (WEKA) data mining framework is selected as a tool to carry out the classification process by using support vector machines and random forests classifiers. Results show that performance of support vector machines is comparable to that of random forests. This study makes a positive contribution to the problem of land cover classification by exploring generalized reduced gradient method, support vector machines, and random forests to improve producer accuracy and overall classification accuracy. The performance comparison of these classifiers is valuable for a decision maker to consider tradeoffs in method accuracy versus method complexity.