APPLICATION OF MACHINE LEARNING IN CREDIT SCORING: PERFORMANCE AND EFFICIENCY COMPARISON OF ENSEMBLE ALGORITHMS

• Main Author: Fikri Nurohman, Muhamad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/83527
• Institution: Institut Teknologi Bandung

Credit plays a crucial role in the economic circulation of a country. Banks and other lending institutions often use stagnant funds to fulfill their role as credit providers to the public or institutions. Before granting credit, creditors typically conduct a credit risk assessment to minimize the potential credit risk. Credit risk assessment can be done through an expert's evaluation or automated models based on the personal data and transaction history of prospective borrowers, known as credit scoring. One method for performing credit scoring is through machine learning. Various algorithms can be used to build a good credit score prediction model. The implementation of machine learning algorithms is carried out in six stages: understanding business requirements, understanding the data used, data preprocessing, modeling and parameter optimization, evaluation, and deployment. Understanding business requirements and data is done through exploratory data analysis. This stage provides insights into the important features to be used in the model-building stage. In the modeling stage, several ensemble learning algorithms are selected for comparison. For each algorithm, parameter optimization is carried out using grid search and k-fold cross-validation. The prediction results from each algorithm will be evaluated using several classification metrics, including AUC, accuracy, F1 score, precision, recall, and training time, with AUC and recall for the "bad" class as the primary metrics.