APPLICATION OF NAIVE BAYES MACHINE LEARNING ALGORITHM FOR CREDIT SCORING IN PEER-TO-PEER LENDING

• Main Author: Annelies Az Zahra, Rifda
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49941
• Institution: Institut Teknologi Bandung

The process of applying for credit at the bank for most people is quite troublesome with the requirements that are not easy, and the disbursement of funds that takes a long time. Financial Technology, especially peer-to-peer (P2P) lending, comes as a quick credit solution with easy submission where most of the procedures are already fully digital. However, the number of non-performing loans from P2P Loans in Indonesia in 2019 reached 3.18%, very high compared to bank NPL which were only 2.77%. Technologies such as machine learning can help the loan application process to be more accurate and efficient. The Naive Bayes algorithm is a fast and simple classification algorithm based on the Bayes probability theorem. In this final project, two types of Naive Bayes algorithms are selected, which are GaussianNB and CategoricalNB. The Solution is implemented using CRIPS-DM method through five stages that includes business understanding, data understanding, data processing, modeling, and evaluation. The optimized model with the GaussianNB algorithm is superior compared to CategoricalNB for the Lending Club dataset with the results of the accuracy of 91.8%, precision 77.5%, recall 89.9%, specificity 92.3%, f1-measure 83.2%, and AUC 95.7%.