IMPLEMENTATION OF MACHINE LEARNING SUPPORT VECTOR MACHINE ALGORITHM IN FRAUD DETECTION SYSTEM

• Main Author: Muhammad Ghifary K, Abidzar
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49866
• Institution: Institut Teknologi Bandung

The rise of the fraud transaction incident requires financial organizations (such as banks) to develop ways to detect fraud transactions to avoid losses occuring from the fraud transaction itself. One of which is machine learning-based fraud detection system. In resolving these problems, the methodology used is CRISP-DM. Various machine learning algorithms have been tried to classify fraud transactions. Support Vector Machine Algorithm is one of the alternative that can be used. The radial based function kernel method is the most appropriate for the characteristics of transaction fraud data compared to other kernel methods. The development of a fraud transaction detection system is carried out in two stages, the first stage is data processing and the second stage is hyperparameter optimization. The results of testing this machine learning model produces performance with accuracy reaching 99.9 %, precision reaching 73 %, textit recall reaching 97.9 %, f1 scores reaching 83.6 %, and the area under textit precison-recall reaching 0.778 and false positive rate reaching 0.01 %