FRAUD DETECTION SYSTEM USING CONVOLUTIONAL NEURAL NETWORK ALGORITHM

• Main Author: Vincent
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/56120
• Institution: Institut Teknologi Bandung

In this modern era, it is undeniable that the internet has become a basic human need which is one of the pillars of the rolling of life. Almost everything humans do today requires an internet connection, one of which is when conducting financial transactions. An example of the use of the internet in financial transactions is the use of mobile banking and digital wallet/electronic money. These two things are indeed very helpful for humans to transact anytime and anywhere very easily and quickly, but behind this ease and convenience there is a threat waiting for an opportunity to attack anyone. This is very possible because the device used is connected to the internet network so there may be intruders from the public network to commit fraud, especially in terms ofmobile banking. In this final project, a machine will be developed to detect fraud (fraud) finance with an artificial intelligence-based approach (artificial intelligence). The machine that will be developed will apply one of the methods in the development of an artificial nervous system, namely Convolutional Neural Networks ( CNN). The development of this machine consists of 5 stages, namely understanding business needs, understanding data, data preparation, machine modeling, and model evaluation. In this study, the performance of the model between before and after optimization will be compared usingframework Optuna who managed to improve the performance of the model from 83% to 91% (recalls).