In this thesis, we deal with the problem of predicting the length of bankruptcy proceedings in Slovenia using different machine learning methods. We have a wide range of data on the proceedings themselves, from 2008 onwards, as well as data on companies and individuals in the proceedings. We started with the preparation of data for statistical analysis, in which we got a good insight into the problem in front of us. Finally, we ended up predicting the lengths of bankruptcy proceedings, finding that the XGBoost model was the best at predicting with MAE of 240 days. Immediately behind it were the random forest and gradient boost models with MAE of 243 days.
|
