Humboldt-Universität zu Berlin - School of Business and Economics

Telephone: +49 30 2093-99543

Email: nikita.kozodoi[at]hu-berlin.de

Office hours: by appointment

Room: SPA1, 335

Short-Bio

Nikita Kozodoi holds a M.Sc. degree in Economics and Management Science from the Humboldt University in Berin. Since April 2018, he works as a researcher at the Chair of Information Systems. As part of his master's thesis, he investigated the predictive performance of managerial heuristics compared to machine learning methods in the area of box office forecasting. In the course of his Ph.D., Nikita researches applications for machine learning in credit risk analytics.

Research Interests

Nikita is particularly interested in machine learning and deep learning applications in credit scoring. His work includes such topics as: developing solutions to mitigate the impact of sampling bias on scoring models; performing profit-driven feature selection with multi-objective algorithms; reducing algorithmic discrimination and analyzing the profit-fairness trade-off.
• Credit risk analytics
• Reject inference
• Semi-supervised learning
• Active learning
• Algorithmic fairness
• Feature selection
• Deep learning

Social Media

Publications

• Kozodoi, N., Jacob, J., & Lessmann, S. (2021). Fairness in Credit Scoring: Assessment, Implementation and Profit Implications. arXiv preprint arXiv:2103.01907. [PDF on arXiv]
• Kozodoi, N., Katsas, P., Lessmann, S., Moreira-Matias, L., & Papakonstantinou, (2020). Shallow Self-Learning for Reject Inference in Credit Scoring. In ECML PKDD 2019 Proceedings (pp. 516-532). Springer, Cham. [PDF on ResearchGate] [published version]
• Kozodoi, N., Lessmann, S., Papakonstantinou, K., Gatsoulis, Y., & Baesens, B. (2019). A multi-objective approach for profit-driven feature selection in credit scoring. Decision Support Systems, 120, 106-117. [PDF on ResearchGate] [published version]
• Kozodoi, N., Lessmann, S., Baesens, B., & Papakonstantinou, K. (2019). Profit-Oriented Feature Selection in Credit Scoring Applications. In Operations Research 2018 Proceedings (pp. 59-65). Springer, Cham. [PDF on ResearchGate]
  [published version]

Packages

• fairness: R package for computing different measures of algorithmic fairness https://github.com/kozodoi/fairness
• dptools: Python package with helper functions for data processing and feature engineering https://github.com/kozodoi/dptools

Personal Website

https://kozodoi.me/