SUMMARY
Main aim of this work was analyzing of the municipality election results and election results of the local self-government bodies. First phase of the work was pre-processing election results for input. The next phase was using of clustering neural network SOM for subsequent processing. And the last and final phase was the interpretation and visualization of the results.
Firstly, data were modified to have the same quantity (6249 villages). The elections which were analyzed were those to European Parliament (2004, 2009), to the lower House of Parliament (2002, 2006) and local elections (2000, 2004, 2008). These elections are analyzed separately for each region (13).
There were two methods of data normalization used in this master’s thesis. The first one is linear transformation . The second one, which achieved better results for the purpose of this thesis, is transformation when max 1 and another is quotient of max . For analysis was used program SOM_AV made in VŠB FEI. For visualization of SOM was used program Java SOMToolbox.
The most important aim of the work was the interpretation of the results. Last step was to visualize neural network into the maps. There were created DBF tables for searching and displaying of the results. There were also created maps which are supplemented with summary tables. Summary tables include the information about results in categories of SOM D-matrix.
The analysis of SOM is one of the possibilities for studying election results. For the local elections are the results inaccurate, that is why the interpretation is rather difficult. The results of this master’s thesis can be useful for political scientists, sociologists and those interested in election results.