Diplomová práce

The thesis deals with the evaluation of crime in the Czech Republic municipalities in the years 2016-2021. The thesis used mainly crime data from the Mapy Kriminality. The crime rate was assessed in a geographical context by calculating the gross and net crime index using geographic information systems. The net crime index as opposed to the gross index includes only the criminal legal population. An innovative and unique approach was the author's design and calculation of a weighted crime index for the year 2021 in the wards of the country. For this purpose, the data of the Police of the Czech Republic of all registered types of crimes according to the tactical-statistical classification was used. The weights included in the calculation were calculated based on the average length of unconditional imprisonment (in months) for each type of crime classified under the Criminal Code. The results of the calculated weighted crime index in the form of maps have a broader explanatory value, can provide new perspectives and come as close as possible to the real state of crime in a given area by taking into account the different weights of crimes. However, it all depends on the quality of the input crime data.

Crime rates are variable in the area. The identification of crime was focused primarily on places showing anomalies and on places with higher or lower concentrations of crime rates. The results of all aggregate crime indices across the years under study show that the highest crime rates in the Czech Republic are primarily most concentrated in Prague, in municipalities in Most, Ostrava, large cities with a high population (50,000 or more), and then in municipalities near the border (with the exception of SE Czech Republic). Other important outputs of the crime rate assessment are animations and matrix maps of crime rate indices of sub-categories of crimes in municipalities of the Czech Republic for the years 2016-2021.

It should be noted that the thesis reflects the need for crime research in a detailed environment, namely at the level of municipalities, which makes this thesis exceptional in the Czech context and unique in terms of the methods and dataset used. The results of the crime rate assessment are somewhat surprising, however, they cannot be compared with other domestic studies, as this is the first study that has been conducted in the context of the data used at the municipal level in the Czech Republic. The map outputs provide useful information on crime trends and patterns in different areas. Mapping the spatial distribution of crime rates can help the Police of the Czech Republic and other law enforcement agencies better understand the causes and develop strategies to address crime.

Both non-spatial linear Ordinary Least Squares (OLS) regression model and spatial regression models - Spatial Lag Model (SLM), Spatial Error Model (SEM) and Geographically Weighted Regression (GWR) - were used to determine the relationship between the net crime index and selected demographic and economic indicators. Regression modelling was carried out for two scenarios - MODEL 1 including municipalities with zero crime index and MODEL 2 not including these municipalities. Although the logarithmic transformation of the crime index data in MODEL 2 contributed to a positive change in the normality of the data, the author's focus was on the interpretation of MODEL 1 to assess a closer approximation of the true crime rate. An evaluation of MODEL 2 would be beyond the scope of this thesis and may represent a potential topic for further research.

Using OLS, it was found that as the proportion of religious population increases, the crime index also increases. In terms of marital status, it is surprising that the model revealed a situation where the crime rate increases as the number of single and divorced people increases. According to the model, as the proportion of the population in insolvency increases, the crime index decreases. The OLS evaluated that as the crime index values increase, the share of residents in foreclosure, the Roma population, and recipients of housing benefits increases.

A partial part of the thesis is devoted to the application of spatial autoregressive models. According to the Akaike Information Criterion (AIC), spatial SLM and SEM models yield a slight improvement over non-spatial models. Since there was no significant improvement in the results compared to OLS, none of these spatial models were chosen for further interpretation.

Using the GWR method led to better results than using the OLS method. Several statistics were used for comparison, including the value of the coefficient of determination R² and AIC. In this case, a higher explanatory power of the GWR model over the OLS model is indicated by a lower AIC value. The width of the fixed Gaussian band was optimized by AIC and R² to 15 km. The resulting model for both scenarios still has problems with high local multicollinearity. In any case, both scenarios did not remove the spatial autocorrelation of the residuals. In interpreting the resulting GWR MODEL 1 regression coefficients, the author's focus is primarily on the statistically significant locations identified by the model. The GWR method was used to search for relationships between the crime index and eight predictors - the age index, the proportion of the population without education, the proportion of the population in foreclosure, the proportion of the population in insolvency, the proportion of the population with religion, the proportion of the population that is single, the proportion of the population that is divorced, and the voter turnout rate).

The results of local GWR regression coefficients can help to understand the specifics of statistical and spatial relationships of crime with respect to selected economic and demographic indicators in the context of geoinformatics and geography. These results can be useful to support crime prevention in specific areas of the country and to improve various measures and strategies. However, it should be borne in mind that crime is a complex and multidisciplinary topic that involves social, psychological and environmental factors. Therefore, the results of the regression coefficients should be interpreted with caution and taking into account other relevant factors. A multidisciplinary approach will allow for a more comprehensive and holistic view of crime in a given area.