<![CDATA[Village Credit Institutions, known locally as Lembaga Perkreditan Desa (LPD), play a fundamental role in Bali's financial inclusion ecosystem by integrating formal credit evaluations with the socio-cultural considerations of indigenous communities. As the volume of credit applications escalates, relying on subjective manual assessments becomes susceptible to inconsistency, potentially triggering non-performing loan risks while simultaneously hindering capital access for potential borrowers. This comparative study empirically evaluates the performance of Support Vector Machine (SVM) and Logistic Regression (LR) algorithms in classifying customer creditworthiness at LPD Sibetan, Karangasem Regency. Utilizing 4,000 historical credit application records from January 2020 to December 2024, both models were extensively optimized using a Grid Search approach with 5-fold cross-validation. The results demonstrate that the SVM model with a linear kernel consistently outperforms Logistic Regression across all evaluation metrics. SVM achieved a classification accuracy of 90.00%, precision of 90.91%, recall (sensitivity) of 96.77%, and an Area Under the Curve (ROC-AUC) score of 98.92%. Conversely, the Logistic Regression model with L1 regularization recorded an accuracy of 86.12% and an ROC-AUC of 97.52%. An anatomy of the confusion matrix reveals that SVM is vastly superior in suppressing both False Positives (financial risk) and False Negatives (opportunity cost). By comparing the robustness of SVM and Logistic Regression, this study contributes to financial inclusion by providing a more accurate credit scoring model for microfinance institutions, ensuring that communities previously deemed high-risk can be reassessed more fairly and gain access to capital.]]>
Copyrights © 2026
