ItemConference Paper

Optimizing Fraud Detection Systems in Credit Card Transactions Using Machine Learning Techniques

Title
Optimizing Fraud Detection Systems in Credit Card Transactions Using Machine Learning Techniques
Creator
Reddy, Veerendra; Rathidevi, T.; Prathap, Boppuru Rudra
Description
Rapid e-commerce services and emerging technologies have grown to use credit card usage as a widespread way of effecting payments, thereby increasing bank transaction volume. It is, therefore, equally increasing fraudulent activitiesthus showing the critical need for fraud detection methods development. Class-weighting hyperparameters are studied and applied to handle class imbalance between fraudulent and legitimate transaction classes. We mainly use Bayesian optimization for these hyperparameters tuning with consideration of unbalanced data problems. The key components of our method involve weight-tuning as a preprocessing step and using the extreme gradient boosting [XGBoost] algorithm to enhance further the light gradient boosting machine [LightGBM] based on an ensemble voting process. Moreover, we use deep learning for hyperparameter tuning with special consideration given to our introduced weight-tuning approach. Experiments on real-world datasets demonstrate the efficiency of our strategies. We follow recall-based metrics and the widely used ROC-AUC scores for the unbalanced datasets, which are more appropriate for measuring the model performance. All the algorithms are compared based on fivefold cross-validation, while the majority voting ensemble method is applied to evaluate the combined performance of the algorithms. The previous results prove that LightGBM and XGBoost perform best, with optimal performances obtained at ROC-AUC scores of 0.95, precision of 0.79, recall of 0.80, and an F1 score of 0.79. Further, deep learning with Bayesian Optimization achieves the ROC-AUC scores of 0.908, precision of 0.96, recall of 0.82, F1 score of 0.88, and Accuracy of 0.9996all of which were significant improvements over the previous approaches. This paper presents Bayesian-optimized LightGBM for fraud detection, where it improves accuracy and efficiently tunes hyperparameters. The main novelty here is integrating Bayesian Optimization into dynamically enhancing model performance for handling class imbalance and reducing false detections. The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.
Source
Lecture Notes in Networks and Systems;Volume;1612 LNNS;pp.339-349
Date
01-01-2026
Publisher
Springer Science and Business Media Deutschland GmbH
Subject
Bayesian optimization; Hyperparameter tuning; Imbalanced datasets; LightGBM; XG Boost
Coverage
Reddy V., Computer Science and Engineering, CHRIST University, Karnataka, Bengaluru, India; Rathidevi T., Computer Science and Engineering, CHRIST University, Karnataka, Bengaluru, India; Prathap B.R., Computer Science and Engineering, CHRIST University, Karnataka, Bengaluru, India
Rights
Restricted Access; Hardcopy may be available in the library
Relation
ISSN: 23673370; ISBN: 978-981952871-4;
Format
online
Language
English
Type
Conference paper

Citation

Reddy, Veerendra; Rathidevi, T.; Prathap, Boppuru Rudra, “Optimizing Fraud Detection Systems in Credit Card Transactions Using Machine Learning Techniques,” CHRIST (Deemed To Be University) Institutional Repository, accessed June 18, 2026, https://archives.christuniversity.in/items/show/25430.

Output Formats