Fraud Detection: binary classification deep learning model training/evaluation project
Built and trained a deep learning fraud-detection model to classify transactions as fraudulent versus genuine. Model performance was evaluated using training and test accuracy metrics to validate classification quality. The work involved using structured transaction features as inputs and verifying outcomes through quantitative evaluation. • Labeled target concept as fraudulent vs genuine (binary classification) for supervised learning. • Engineered/used features such as amount, frequency, and device information for model inputs. • Reported training accuracy of 95.02% and test accuracy of 95.10% as evaluation results. • Implemented the modeling and evaluation workflow in a Python-based ML stack.
2021 - 2025