COMPARISON OF CLASSIFICATION MODELS USING SUPPORT VECTOR MACHINE (SVM) AND K-NEAREST NEIGHBOR (K-NN) METHODS IN NON-PERFORMING LOAN ANALYSIS

Abstract

SVM works by finding the best dividing line (or hyperplane) to separate two groups of data based on the maximum margin.k-NN classifies new data based on its similarity to previous, already-labeled data points.Non-performing loan analysis is a crucial aspect of credit risk assessment. This study compares the performance of the Support Vector Machine (SVM) and k-Nearest Neighbor (k-NN) classification methods in analyzing non-performing loans at PT. Adira FinanceAmbon Branch. The dataset includes demographic and financial attributes, processed through normalization, data splitting, and evaluation using accuracy, precision, recall, and AUC metrics. The results show that SVM with a linear kernel performs best, achieving 97.83% accuracy and 95% AUC. Meanwhile, k-NN with k=5 attains 78.26% accuracy. Thus, SVM outperforms k-NN in classifying non-performing loans in this study.