DESIGN AND IMPLEMENTATION OF BAD DEBT PREDICTOR AND FRAUD MANAGEMENT SYSTEM USING MULTILAYER NEURAL NETWORK

• Main Author: Setiadi, Wiwit
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82921
• Institution: Institut Teknologi Bandung

One of the main problems in telecommunications business, especially in postpaid service at service provider side, is unpaid invoice or bad debt. As the essential part of its cycle, an operator must be able to handle this problem since it contributes significantly in its earnings. Abuse to service is one of the causes resulting bad debt. This kind of fraud is difficult to be managed if the fraudster has entered the network. But, by learning the pattern recognition of the usage, this disservice can be minimized. It depends on the selection of method that is able to recognize the usage pattern. The importance of this problem now becomes a challenge. The challenge is in the form of finding correct concept or method to detect fraud. Some researchers have been address this problem by using various technologies of data mining and intelligent composing. Artificial neural network (neural network) is one of the technologies which will be employed, where its ability to recognize pattern is a pre-eminent value. On this thesis, we design and implement a Bad Debt predictor and Fraud Management System (FMS). This system uses artificial neural network technology. This system is intended be to be used for postpaid services in cellular telecommunications business. Nevertheless, similar concept can also be applied into other technology with a few modification. This research has given good result in the ability of bad debt prediction to 87%, while for FMS has reached 85%.