Penelitian ini bertujuan untuk mengetahui teknik pemodelan probability of default yang tepat pada portfolio pembiayaan bersama antara bank dan fintech lending dan perusahaan pembiayaan (multi finance) pada Bank Swasta Nasional. Teknik yang digunakan sebagai perbandingan adalah teknik tradisional regresi logistik dan teknik machine learning. Penelitian dilakukan berdasarkan data Bank Swasta Nasional dari April 2019 hingga Desember 2022 dan diklasifikasikan menjadi dua kategori yaitu defaiult dan non-default sehingga model dapat melakukan prediksi kemungkinan nasabah gagal bayar selama durasi pinjaman. Analisis dilakukan menggunakan data aplikasi dan data transaksi tanpa melanggar aturan menjaga kerahasiaan nasabah. Selanjutnya, analisis dilakukan hanya untuk produk pembiayaan bersama pada fintech lending dan multi finance yang beroperasi hanya di Indonesia. Variabel signifikan yang memengaruhi pemodelan probability of default pada pembiayaan bersama antara bank dan fintech lending/multi finance antara lain adalah tenor, tujuan pinjaman, nominal beban bunga, deskripsi pekerjaan, kota domisili, umur dan pendapatan. Lebih lanjut, teknik tradisional regresi logistik memiliki tingkat keandalan lebih tinggi dibandingkan pendekatan machine learning decision tree pada studi kasus BSN.

---

This research is to find the appropriate credit scoring model technique to build the default model based on the joint financing schemes product between bank and fintech lending and multi finances that conducted by Bank Swasta Nasional. The credit scoring model to be compared using traditional approach, logistic regression against machine learning technique. This research is case study in Bank Swasta Nasional’s portfolio starting April 2019 up to December 2022 and will be classified into default or non-default so the model can predict the possibility of customer default during the period. The analysis conducted based on variables from application and transaction data that not breaching the confidentiality of personal data in Bank Swasta Nasional. Furthermore, the analysis only applicable for joint financing schemes product to fintech lending and multi finances that already have operated in Indonesia only. The significant variable to model the probability of default for joint financing schemes between bank and fintech lending or multi finances are tenure, loan purpose, interest amount, job description, home city, age and declared income. Furthermore, the analysis showed that the traditional technique logistic regression had higher accuracy compared to machine learning using decision tree in the case study.