Abstract:

Debt collection companies buy overdue debts on the market in order to collect them and recover the highest possible amounts of debt. The key element of purchasing debt portfolios is their valuation at the time of purchase. The article proposes a bulk debt valuation model based on decision trees and binary classification of debts into two classes: at risk of default and fully recoverable. Such division into two extreme classes corresponds to the actual distribution of the value of the recovered bulk receivables. The effectiveness of detecting receivables at risk of default has turned out to be satisfactory, while the accuracy of the classification of receivables that can be repaid by debtors is medium. Other machine learning algorithms also showed a similar quality of classification. The advantage of the decision tree is the ability to generate decision rules that could be used by experts in the process of valuation and making decisions regarding the further treatment of receivables. The proposed approach allows for the elimination of debts that are difficult to recover at the initial stage of valuation.