Text
Metode Generalized Mean Distance-based k-Nearest Neighbor Classifier (GMDKNN) untuk Analisis Credit Scoring Calon Debitur Kredit Tanpa Agunan (KTA)
ABSTRAK
Kredit Tanpa Agunan (KTA) adalah salah satu fasilitas kredit yang disediakan bank, dimana calon debitur dapat meminjam sejumlah dana dari bank tanpa harus memberikan jaminan atau agunan. Credit scoring adalah proses yang bertujuan untuk menilai kelayakan permohonan kredit serta mengklasifikasi para pemohon kredit ke dalam calon debitur yang permohonan kreditnya layak untuk diterima dan calon debitur yang permohonan kreditnya sebaiknya ditolak. Salah satu metode statistika yang dapat diterapkan dalam mengkaji analisis credit scoring adalah metode klasifikasi Generalized Mean Distance-Based k-Nearest Neighbor (GMDKNN). Kajian empiris pada metode ini menggunakan 23.337 data calon debitur KTA pada tahun 2018, dengan variabel dependen yaitu keputusan akhir credit scoring dan tujuh variabel independen yaitu usia, tanggungan anak, lama bekerja, lama perusahaan, pendapatan, pinjaman yang diajukan, dan durasi kredit. Berdasarkan uji seleksi fitur, semua variabel independen dinyatakan berpengaruh secara signifikan pada keputusan akhir credit scoring. Model klasifikasi terbaik diperoleh pada parameter k = 137 dan p = -1 dengan ukuran performa klasifikasi yang dinyatakan dengan nilai APER = 0,2580, akurasi = 74,20%, sensitivitas = 0,6083, spesifisitas = 0,8393, AUC = 0,7238, dan G-Mean = 0,7146.
Kata Kunci: Kredit Tanpa Agunan (KTA), credit scoring, klasifikasi, Generalized Mean Distance-Based k-Nearest Neighbor (GMDKNN).
ABSTRACT
Unsecured Credit is one of the credit facilities provided by banks, where the prospective debtor can borrow some amount of fund from the bank without having to provide collateral. Credit scoring is a process that aims to assess the worthiness of credit applications and classify the credit applicants into prospective debtors whose the credit application is worthy to be accepted and prospective debtors whose the credit application should be rejected. One of the statistical methods that can be applied in examining the analysis of credit scoring is the Generalized Mean Distance-Based k-Nearest Neighbor (GMDKNN) classifier. Empirical study on this method uses 23,337 data of prospective debtor of unsecured credit in 2018, with the dependent variable being the credit scoring final decision and seven independent variables, i.e. age, child dependent, length of employment, age of the company, income, loan proposed, and duration of credit. Based on the feature selection test, all independent variables are significantly taking effect on the credit scoring final decision. The best classification model is obtained in the parameters k = 137 and p = -1 with the classification performance metrics represented by the values of APER = 0,2580, accuracy = 74,20%, sensitivity = 0,6083, specificity = 0,8393, AUC = 0,7238, and G-Mean = 0,7146.
Keywords: Unsecured Credit, credit scoring, classification, Generalized Mean Distance-Based k-Nearest Neighbor (GMDKNN).
706E19III | 706 E 19-ii | Perpustakaan FSM Undip (Referensi) | Tersedia |
Tidak tersedia versi lain