The Corruption Perception Index: analysis of dependence on socio-economic indicators

The article presents the results of applying data analysis methods, in particular cluster analysis and machine learning methods for corruption analysis. The analysis of dependence of the corruption perception index on social and economic indicators in different countries of the world was carried out. Data was collected for the formation of the feature space, the most significant features that have the greatest impact on the corruption perception index were selected, and the countries of the world were clustered according to the selected features. A scheme for conducting a comprehensive analysis to identify the most significant signs and causes affecting the corruption perception index has been developed. Based on the results of clustering, a classification model that can predict the level of corruption in the country based on the values of selected attributes was trained. It is based on the composition of Bagging algorithms with decision trees as basic classifiers. The results of the study can be used by public authorities to develop, organize and adopt appropriate counteraction measures. A software tool based on in the Python language, which allows to perform appropriate analysis using updated data was developed.