Climatic variable selection using random forests regression for malaria transmission modelling in Mpumalanga Province, South Africa.
| dc.contributor.advisor | Gebreslasie, Michael Teweldemedhin. | |
| dc.contributor.author | Kapwata, Thandi. | |
| dc.date.accessioned | 2016-09-01T09:31:11Z | |
| dc.date.available | 2016-09-01T09:31:11Z | |
| dc.date.created | 2015 | |
| dc.date.issued | 2015 | |
| dc.description | Master of Science in Environmental science. University of KwaZulu-Natal, Durban 2015. | en_US |
| dc.description.abstract | Malaria is one of the wold’s most prevalent vector borne diseases with sub-Saharan Africa bearing the highest burden of reported cases. Climate is one of the major determinant factors of malaria transmission as it influences the spatial and temporal pattern of transmission. It is therefore important to be able to understand the relationship between climatic variables and malaria transmission because an understanding of the interactions between them at a local level is an important part in potential outbreaks, targeting vector control strategies, and developing malaria early warning systems. This study covered the Ehlanzeni district of Mpumalanga province in South Africa. It was aimed at determining the climatic variable importance of temperature, lag temperature, rainfall, lag rainfall, humidity, altitude and NDVI in relation to malaria transmission. The random forest algorithm was used to relate the climatic variables extracted from remote sensing imagery and malaria case data collected from health facilities in order to establish individual measures of variable importance and to develop a spatial and temporal prediction models. In this study altitude appeared to be the most responsible variable for malaria transmission because it was most frequently selected as one of the top variables with the highest variable importance followed by NDVI and temperature. The combination of climatic variables that produced the highest coefficient of determination values was altitude, NDVI, and temperature. This suggests that these 3 variables have high predictive capabilities and as a result they should be selected for spatial and temporal modelling of malaria. Furthermore, it was expected that the predictive models generated by the random forest algorithm could be used as an operational malaria early warning system using forecast climatic variable identified in this study in order to assist in containing any potential reoccurrence of malaria after elimination. | en_US |
| dc.identifier.uri | http://hdl.handle.net/10413/13324 | |
| dc.language.iso | en_ZA | en_US |
| dc.subject | Malaria--South Africa--Mpumalanga. | en_US |
| dc.subject | Malaria--Transmission--South Africa--Mpumalanga. | en_US |
| dc.subject | Malaria--Environmental aspects--South Africa--Mpumalanga. | en_US |
| dc.subject | Epidemiology. | en_US |
| dc.subject | Algorithms. | en_US |
| dc.subject | Theses--Environmental science. | en_US |
| dc.subject | Climate. | en_US |
| dc.subject | Forest regression. | en_US |
| dc.title | Climatic variable selection using random forests regression for malaria transmission modelling in Mpumalanga Province, South Africa. | en_US |
| dc.type | Thesis | en_US |