ARTIFICIAL NEURAL NETWORK USING SPIRAL OPTIMIZATION ALGORITHM TO PREDICT CREDIT DEFAULT CUSTOMER
Credit default can be defined as the borrower’s failure to make loan payments on the due date. Credit default can cause losses for lenders, so preventive actions must be taken, one of which is to predict the potential for default early. This credit default customer problem can be categorized as a bi...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/49712 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Credit default can be defined as the borrower’s failure to make loan payments on the due date. Credit default can cause losses for lenders, so preventive actions must be taken, one of which is to predict the potential for default early. This credit default customer problem can be categorized as a binary classification problem. Artificial neural networks (ANN) are a widely used model for classification problems. ANN has the capability to capture linear and non-linear trends from complex data and able to obtain reliable predictions for new data. The backpropagation learning algorithm is a learning algorithm that is widely used in ANN and can provide good results. However, this method has a drawback that is could be trapped in local optima since they perform trajectory searching and require gradient information in the process so that only differentiable functions can be used. In this final project, the author uses a spiral optimization algorithm introduced by Tamura and Yasuda (2011). This method is one of the metaheuristic optimization methods in which the process of finding a solution is not based on trajectory searching and does not require gradient information as learning algorithm for ANN. |
|---|