#TITLE_ALTERNATIVE#

• Main Author: JOSEPHINE NAOMI (NIM: 18214021), JESSY
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28276
• Institution: Institut Teknologi Bandung

<p align="justify"> <br /> <br /> The financial industry is currently facing several problems, including non-performing loans and lack of credit sales. Clustering can help alleviate this problem by improving the performance of the credit scoring classification and segmenting the debtor to subsequently improve marketing. One of the most commonly used clustering algorithms is K-Means. This final project is completed with CRISP-DM methodology. The final result shows that the best model for the SIJEKH dataset is the K-Means algorithm with the value of 'k-means ++' for the init parameter, 10 for the n_init parameter, 0.00001 for the toll parameter, 1 for the n_jobs parameter, 'elkan' for algorithm parameters, 78 for random_state parameter, and default for parameter max_iter, precompute_distances, and copy_x. This model gives best result of segmentation for cluster number of two clusters. Cluster 1 consists of debtors with a smaller nominal income, longer duration, lower nominal installment, and have a greater percentage of problem loans than cluster 2. While cluster 2 consists of debtors with larger nominal income, shorter duration, higher nominal installments, and lower percentage of problem loans compared to clusters 1. Comparison of classification results generally indicates an increase after segmentation for three types of classification, i.e. Decision Tree, K-Nearest Neighbors, and Support Vector Machine. Further research should be done to get the most appropriate business decision to increase credit sales in each cluster. In addition, the scorecard for each cluster should also be done to reduce the risk of non-performing loans for each cluster in the future. <br /> <p align="justify">