Data analytics driven quadratic optimization for solving agent-debtor assignment in debt recovery problem

Date

2025-8-20

Author

Karabaş, Toygun

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

820
views

0
downloads

Cite This

Late payments and irrecoverable debts are rising in today's economy, and proper management is crucial for financial stability. An effective debt collection process begins with the analysis of historical customer data to identify payment trends. The key challenge is to assign the most suitable agents to debtors for recovery. This work proposes a novel hybrid optimization framework which is composed of two phases. In the first phase, a machine learning-based predictive analytics pipeline is proposed to predict the debt recovery rate of all agent-debtor pairs. The second phase addresses the problem of grouping agents and debtors simultaneously and matching those groups with each other in a way that maximizes recovery efficiency. This problem is formulated as a constrained binary optimization problem that proposes quadratic and linear mathematical models. To solve these mathematical models in a reasonable time, three optimization approaches are proposed. This hybrid optimization framework is tested with synthetic and mimicked data.

Subject Keywords

Data analytics, Generative adverserial networks, Iterated local search, Constrained optimization, Debt recovery optimization

URI

https://hdl.handle.net/11511/115647

Collections

Graduate School of Natural and Applied Sciences, Thesis