GIS-baserad prediktion av HIV : en förstudie

Sammanfattning: Epidemic Human Immunodeficiency Virus (HIV) due to its rapid spread and deep influence has been a unique phenomenon in the near history. The virus has been existing all over the world, the spread of infection is both dynamic and complex. Epidemics are a geographical phenomenon with a certain extent. Most of the factors that can be linked to the epidemic are also geographically dependent. Considering that, the Geographic Information System (GIS) can be an important tool in studying the phenomenon.The pilot study spatially predicted HIV level and investigated to what extent this is possible and how the process may look like. The focus of the study was on the technical part, to evaluate the method. The pilot study mainly used the GIS-tool Geographically Weighted Regression (GWR), which is based on regression analysis. The regression type that was used was multiple linear regression that uses a dependent variable and several explanatory variables. The method assumes that the dependent variable can be explained by several variables that have a linear relationship to the dependent variable.The study area of the pilot study is Tanzania, a country located in eastern Africa. Tanzania has large socio-economic gaps and varying levels of HIV between regions. HIV data used in the pilot study comes from the Tanzania HIV Impact Survey (THIS) and the data for other variables comes from The Demographic and Health Surveys (DHS) Program. The HIV model is explained on the basis of four factors: undernourished children, circumcised men, and people with only primary education and households that own agricultural land.The first step in the implementation was data preparation, the next step was to conduct a global analysis using the Ordinary Least Squares (OLS) method. Followed by a geographic analysis with the GWR tool. The last step was to implement predictions using the created model. Four control regions in Tanzania were predicted, the model was also tested in five other African countries. The OLS analysis generates an Adjust R-Squared value of 0,436 which is a measure of the model's explanatory value of HIV. The same value for the GWR analysis was 0,502. The transition to GWR provided an improvement. Two out of four control regions in Tanzania provide accurate results. The same model also predicts relatively well in other African countries. The pilot study's results are limited by low data resolution and limited identification of HIV factors. With higher data resolution and better assurance of HIV factors, more accurate and detailed HIV predictions can be achieved. The pilot study's conclusion is that HIV predictions that obtain credible results are possible with the help of GIS-based analysis tools. As a suggestion for improvement, more high-resolution data is recommended, preferably as points. This would give the analysis better conditions for more accurate and detailed predictions.