Semi Supervised Learning Model Untuk Prediksi Kelancaran Pembayaran Kredit Kendaraan Bermotor

Abstrak

The high demand for motorcycle credit has led to an increase in the number of financing companies
to help people or consumers buy motorcycles on credit. However, there are challenges in vehicle
credit payments where most customers often do not comply with installment payment deadlines,
some even try to avoid their payment obligations which have a negative impact on the company's
finances. To overcome this problem, an analysis was made using a semi-supervised learning
technique that is able to predict the extent to which vehicle credit payments are smooth by
customers. This study uses 2 algorithms, namely the K-Means Clustering algorithm and the
Random Forest algorithm. The K-Means Clustering algorithm is used to create target labels for
classification analysis testing. From the results of the analysis, k = 2 was obtained, with an SSE
value of 62.23 and a Silhouette Score value of 0.53. So that each cluster can be named, namely
smooth and stalled clusters. After the label is known, the Random Forest classification analysis is
carried out. There are several features that greatly influence the smooth payment of motor vehicle
credit such as
‘PEKERJAAN’,’PENGHASILAN_PER_BULAN’,’JUMLAH_TANGGUNGAN’ and ‘CICILAN_NASABAH. The analysis results showed an accuracy score of 98%, a precision of
100%, a recall of 97%, and an F1 Score of 98%.