<![CDATA[Assessing sharia financing eligibility is a critical process for Islamic microfinance institutions such as Baitul Maal wat Tamwil (BMT), as inaccurate financing decisions may increase financing risk and affect institutional sustainability. However, financing evaluations are often conducted manually and rely heavily on subjective judgment, leading to inconsistencies and potential bias in decision-making. Therefore, this study aims to evaluate the effectiveness of Correlation-Based Feature Selection (CFS) in identifying relevant financing attributes and to compare the classification performance of the Naïve Bayes and Decision Tree (J48) algorithms for sharia financing eligibility assessment. The dataset used in this study consists of 500 historical financing records obtained from BMT Indragiri, comprising 127 eligible and 373 ineligible financing applications. The research process included data preprocessing, feature selection using CFS with the BestFirst search strategy, model construction, and classification using WEKA 3.8. Model performance was evaluated using 10-fold cross-validation based on accuracy, precision, recall, F1-score, Kappa statistic, and Area Under the Curve (AUC). The feature selection results showed that all predictor attributes, namely income, number of dependents, employment status, and financing history, were retained by the CFS algorithm, indicating that each attribute contributes relevant information to financing eligibility classification. Experimental results revealed that the Naïve Bayes classifier achieved an accuracy of 92.4%, precision of 92.3%, recall of 92.4%, F1-score of 92.2%, Kappa statistic of 0.7896, and AUC of 0.971. Meanwhile, the Decision Tree (J48) classifier achieved superior performance with an accuracy of 95.6%, precision of 95.7%, recall of 95.6%, F1-score of 95.6%, Kappa statistic of 0.8851, and AUC of 0.957. In addition, the Decision Tree model generated 23 decision rules and a tree size of 42 nodes, providing transparent and interpretable knowledge to support financing eligibility assessment. The findings indicate that the Decision Tree (J48) algorithm outperformed Naïve Bayes in classifying sharia financing eligibility and offers the additional advantage of interpretable decision rules. The proposed approach can support more objective, consistent, and transparent financing decision-making processes in Islamic microfinance institutions.]]>
