IMPLEMENTATION OF ARTIFICIAL NEURAL NETWORK ALGORITHM IN FRAUD DETECTION SYSTEM

• Main Author: Bagus Raditya A.M, Ida
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75677
• Institution: Institut Teknologi Bandung

Digital technology is becoming increasingly popular and is being adapted into various economic processes. The rapid development of the digital economy also has a significant impact on the financial sector. This development has driven the transition from conventional payment methods to digital payments. One popular method of digital payment is mobile payment using credit cards. However, the risk of fraud and fraudulent activities has also been on the rise, with rising numbers of fraud in credit cards transaction. In this final project, a fraud detection system will be built, implementing an Artificial Neural Network algorithm using the Python programming language. The goal of the fraud detection system is to classify credit card transactions as either fraudulent or legitimate. The development of the fraud detection system will follow the CRISP-DM (Cross- Industry Standard Process for Data Mining) methodology. The fraud detection system will be trained using the IEEE-CIS Fraud Detection dataset. In the Business Understanding stage, the aim is to create a fraud detection system that can detect as many fraudulent transactions as possible. In the Data Understanding stage, exploratory data analysis will be performed. Data Preparation will involve data processing with imputation, feature selection, and data sampling. The Modeling stage will focus on building the fraud detection model using the Tensorflow library and utilizing Optuna for hyperparameter tuning. Finally, the Evaluation stage will assess the performance of the fraud detection system using several evaluation metrics, such as False Positive Rate (FPR), Area Under the Curve (AUC), and recall. The final ANN model achieved a performance with an FPR of 0.311, AUC of 0.817, and recall of 0.776.