SOCIAL MEDIA DATA TO IMPROVE CREDIT SCORING ACCURACY WITH A DATA MINING APPROACH BASED ON SUPPORT VECTOR MACHINE: CASE STUDY OF AN ONLINE PEER TO PEER LENDING IN INDONESIA

• Main Author: Saputra, Okta
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/49695
• Institution: Institut Teknologi Bandung

In recent years, financial technology (fintech) is rapidly expanding in Indonesia. The fintech ecosystem in Indonesia is dominated by Peer to Peer (P2P) lending. Small and micro-enterprises and individual borrowers do not need loan guarantors and collateral in getting the financing. Yet, this condition will position P2P Lending to be exposed with credit risk. The output of data mining technique can be used to make the credit scoring model better, one of the algorithms by using Support Vector Machine. There is one online peer to peer lending company in Indonesia developing social evaluation to enhance their credit scoring to face the fluctuation of non-performing loan. Therefore, the aim of this study is to build credit scoring model using social media data based on Support Vector Machine. The model development process adapted Cross-Industry Standard Process for Data Mining (CRISP-DM), which consists of business understanding, data understanding, data preparation, model development, and evaluation. The borrower’s data in the company is used including the borrower’s demographic, historical payment data, and social media data. The research that has been done resulted that SVM Linear has the best performance compared to other kernels. By adding social media data, it can increase the performance of the credit scoring model measured in AUC as much as 7%. To build the credit scoring model more efficient, the company can simply use 8 features which are tenor, posting_midnight, following, igpost_month, employment_karyawan, employment_wiraswasta, ig_month, employment_other, and income.