Abstract

Traditional credit assessment methods rely on individuals' credit history and their interactions with financial institutions. However, these methods are insufficient for individuals, leading to limitations in financial inclusion. The integration of alternative financial data sources enables more comprehensive and accurate credit risk predictions. This study will examine the role of mobile payment history, bill payments, and e-commerce behavior in credit risk assessment. Open data sources will be utilized to enhance financial inclusion and improve credit evaluation processes.

In this study, the World Bank Global Financial Inclusion Database, the Brazil Open Data Portal, and the UK Open Banking API will be used as data sources. The World Bank database will be employed to analyze financial accessibility. The Brazil Open Data Portal will provide comprehensive insights into e-commerce behavior, while the UK Open Banking API will supply extensive data on digital banking transactions.

To process these data sources, Logistic Regression, Decision Tree, and XGBoost algorithms will be used. Logistic Regression will be applied to provide interpretable results for binary classification tasks. Decision Tree will be used to better understand dataset structures and efficiently process information from alternative data sources. XGBoost will be used to achieve high accuracy in large-scale datasets. As a result of this study, we aimed to analyze alternative credit score evaluation methods with Machine Learning.