Optimization of the assignment problem in debt enforcement considering socio-economic aspects of the debtor

The main objective of this thesis was to design a model for an economically efficient assignment of enforcement cases to active bailiffs, while fulfilling requirements for workload distribution and considering the socioeconomic aspects of the debtor. The theoretical part is based on a review of academic literature and public sources regarding enforcement proceedings in the Czech Republic, operations research, and machine learning methods. The practical part utilizes anonymized data provided by a private entity, the representativeness of which is evaluated through comparison with publicly available data from the Czech Map of Indebtedness. An exploratory data analysis is conducted on these data, establishing key premises for the subsequent modeling. Based on the processed data, the expected recovery rate of individual claims is predicted using a Random Forest algorithm. The resulting predictions are integrated into an optimization model formulated as a generalized assignment problem. The implementation utilizes an exact integer linear programming algorithm. Two versions of the resulting model are proposed: the first addresses the allocation of claims while respecting uniform capacity constraints among individual bailiffs, whereas the second integrates specific geographical constraints. The results suggest that, within the framework of the defined environment, both versions of the proposed solution demonstrate higher efficiency than the current assignment method, while considering selected procedural aspects affecting the debtor through proxy penalty indicators.