Measuring and Mitigating Bias in Bank Customers Data with XGBoost, LightGBM, and Random Forest Algorithm
DOI:
https://doi.org/10.26555/jiteki.v9i1.25768Keywords:
Bank customers data, Bias, Fairness, Bias mitigation, Bias detectionAbstract
To retain its clients, the Portuguese banking institution conducts direct marketing in the form of phone calls to conduct marketing so that clients subscribe to the bank's term deposit. The data used is named bank customers data. Important client features are considered in the acquisition process. This research was conducted with bank customers data from Portuguese banking institution which implements agent acquisition. With a large number of data on bank customers, it can lead to a diversity of data which allows the results of agent acquisition to be unfair. With this, a bias detection and mitigation algorithm are needed to achieve fairness. AI fairness 360 (AIF 360) is a toolkit that provided a bias detection and mitigation algorithm. The bias mitigation algorithm in AIF 360 is divided into three processes, namely reweighing and learning fair representation at the pre-processing stage, prejudice remover and adversarial debasing at the in-processing stage, and equalized odds and reject option classification at the post-processing stage. The output of this study is a comparison of the calculation of bias detection with disparate impact (DI) and statistical parity differences (SPD) before and after mitigation. The adversarial debiasing algorithm performed best than others with 0.943 of DI, -0.004 of SPD, and also increased the 0.015% of the AUC score. Conducting this research can help the prediction of client’s term deposits in Portuguese banking institution more fairly.
Downloads
Published
Issue
Section
License
Authors who publish with JITEKI agree to the following terms:
- Authors retain copyright and grant the journal the right of first publication with the work simultaneously licensed under a Creative Commons Attribution License (CC BY-SA 4.0) that allows others to share the work with an acknowledgment of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgment of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work.
This work is licensed under a Creative Commons Attribution 4.0 International License