Comparison of cluster analysis methods using data of votes of Chamber of Deputies of the Parliament of the Czech Republic

The aim of this diploma thesis is to compare methods of hierarchical cluster analysis for nominal data. These methods differ in the similarity measures used and are evaluated using several evaluation criteria. For their practical comparison, data on the voting of members of the Chamber of Deputies of the Parliament of the Czech Republic for the period 2013–2017 were used. According to specific criteria, four subsets of votes are selected from the available voting records, which are used to prove the aggregation of members. Thus, in addition to the comparison of methods of cluster analysis, the work obtains information about the identified clusters in the Chamber of Deputies, about the behaviour of individual deputies and whether the unity of political parties or governing coalition is preserved. The theoretical part of the thesis is devoted to the introduction of cluster analysis with emphasis on clustering of categorical data. Hierarchical clustering and similarity measures between objects are discussed in more detail. In the practical part, the analysis of subsets is performed, where the combinations of the number of clusters and the similarity measure used, which are the most suitable according to the evaluation criteria, are examined in more detail. The conclusion of the thesis is the identification of similarity measures, which application led to the best or the worst results. The factual conclusion is the prevailing observance of the uniform vote of government coalitions and political parties, where the inclusion of members of the Communist Party and the Úsvit movement to some of the groups of other political parties is particularly variable.