Integrasi Metode Analytic Hierarchy Process (AHP) Dengan Machine Learning Untuk Memprediksi Tingkat Risiko Dan Klasifikasi Pinjaman Sektor UMKM Di Bank X

Riskandy, Yudi Hendra (2025) Integrasi Metode Analytic Hierarchy Process (AHP) Dengan Machine Learning Untuk Memprediksi Tingkat Risiko Dan Klasifikasi Pinjaman Sektor UMKM Di Bank X. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6047231044-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Prediksi risiko kredit yang akurat dan transparan merupakan tantangan penting dalam sektor perbankan, khususnya untuk pembiayaan usaha mikro, kecil, dan menengah (UMKM). Model prediktif konvensional kerap bersifat kaku dan kurang adaptif terhadap karakteristik lokal serta intuisi pakar. Penelitian ini menawarkan pendekatan hybrid melalui integrasi Analytic Hierarchy Process (AHP) dan Machine Learning (ML), guna membangun sistem klasifikasi risiko kredit yang lebih fleksibel, dapat dijelaskan, dan relevan dengan praktik perbankan nasional. Metode AHP digunakan untuk menangkap penilaian pakar terhadap empat fitur utama risiko pinjaman—installment, debt-to-income ratio (DTI), loan amount, dan annual income—yang kemudian dihitung bobotnya melalui perbandingan berpasangan. Bobot ini digunakan untuk menyesuaikan ambang klasifikasi (threshold) model prediktif Random Forest secara dinamis, menggunakan rumus: Thresholdᵢ = 0.5 × (1 + Skor AHPᵢ). Model dievaluasi menggunakan metrik akurasi, precision, recall, F1 score, ROC AUC, dan Cohen’s Kappa. Random Forest menunjukkan kinerja tertinggi (akurasi 90,7%, F1 score 91%, AUC 96,5%). menyesuaian threshold berbasis AHP terbukti meningkatkan presisi klasifikasi pinjaman bermasalah (bad loans), dengan tetap menjaga keseimbangan performa. Temuan ini menunjukkan bahwa integrasi AHP tidak hanya memperkuat aspek teknis model, tetapi juga menjembatani sistem prediksi dengan pertimbangan manajerial dan regulasi. Model yang dihasilkan lebih adaptif terhadap konteks kebijakan risiko bank dan mampu meningkatkan akuntabilitas serta efektivitas pengambilan keputusan kredit berbasis data dan pengetahuan pakar.
=======================================================================================================================================
Accurate and transparent credit risk prediction remains a critical challenge in the banking sector, particularly in financing micro, small, and medium enterprises (MSMEs). Conventional predictive models often lack flexibility and fail to incorporate expert judgment, which is essential in dynamic financial contexts. This study proposes a hybrid approach by integrating the Analytic Hierarchy Process (AHP) with Machine Learning (ML) to develop a credit risk classification system that is more adaptable, interpretable, and aligned with local banking practices. AHP was utilized to quantify expert assessments of four key loan risk features—installment, debt-to-income ratio (DTI), loan amount, and annual income—by means of pairwise comparisons. The resulting weights were used to dynamically adjust the classification threshold in a Random Forest predictive model, applying the formula: Thresholdᵢ = 0.5 × (1 + AHP Scoreᵢ). The model was evaluated using accuracy, precision, recall, F1 score, ROC AUC, and Cohen’s Kappa. Among the four ML algorithms tested (Random Forest, J48, Bagging, LMT), Random Forest achieved the best performance with 90.7% accuracy, 91% F1 score, and 96.5% AUC. After incorporating AHP-based threshold adjustments, the model demonstrated improved precision in identifying bad loans, while maintaining balanced overall performance. The findings suggest that AHP integration enhances not only the technical robustness of the model but also its alignment with managerial judgment and regulatory requirements. The resulting model is better suited to risk-based credit policies and provides a more accountable, data-informed decision support system that harmonizes algorithmic prediction with domain expertise.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Machine Learning, Analytic Hierarchy Process (AHP), Risiko Kredit, UMKM, Threshold Adaptif, Random Forest, Explainable AI Machine Learning, Analytic Hierarchy Process, Credit Risk, MSMEs, Adaptive Threshold, Random Forest, Explainable AI
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Yudi Hendra Riskandy
Date Deposited:	30 Jul 2025 03:40
Last Modified:	30 Jul 2025 03:40
URI:	http://repository.its.ac.id/id/eprint/122764

Actions (login required)

View Item