Application of Data Mining for Predicting Bank Customer Credit Eligibility Using the K-Nearest Neighbor (KNN) Algorithm

Abstract

Credit granting is a core banking service that carries a high risk of non-performing loans if thecustomer eligibility assessment process is not conducted accurately. Manual evaluation processestend to be time-consuming and are potentially influenced by the subjectivity of credit analysts.Therefore, a data-driven approach is needed to support decision-making that is more objective,rapid, and consistent. This study aims to apply data mining techniques using the K-NearestNeighbor (KNN) algorithm to predict customer credit eligibility based on historical loanapplication data. The research methodology follows the Knowledge Discovery in Databases(KDD) process, comprising data selection, data preprocessing, data transformation, data mining,and evaluation. The dataset used is the "Loan Prediction Dataset" from Kaggle, consisting of 614records and 12 attributes, processed using RapidMiner Studio. The transformation stage involvedlabel encoding for categorical attributes and normalization for numerical attributes beforesplitting the dataset into 80% training data and 20% testing data. The KNN model wasconstructed using the Euclidean Distance metric and evaluated using a confusion matrix basedon accuracy, precision, and recall. The results indicate that the model achieved an accuracy of74.80%, a precision of 75.31%, and a recall of 74.80%. These findings demonstrate that the KNearest Neighbor algorithm delivers satisfactory classification performance, making it aviable alternative for supporting objective, data-driven decision-making in the credit eligibilityassessment process.