Skill-Driven Loan Prediction for Economic Inclusion: A Computational Approach for Financial Empowerment

In this paper, we propose an innovative loan prediction framework by developing an Adaptive random forest ensemble (ARFE) model, that integrates financial metrics with skill-based insights derived from the GEMMA large language model (LLM). Traditional credit evaluation models often overlook applicants’ skills, potentially excluding capable individuals with limited financial histories. The proposed ARFE model adapts to both financial and skill attributes, leveraging GEMMA’s fine-tuned skill ratings to enhance prediction accuracy and inclusivity. By emphasizing applicants’ human skill, ARFE offers a more equitable approach to loan evaluation, particularly benefiting underrepresented individuals in resource-limited regions. Comparative analysis with other baseline models demonstrates ARFE’s superior performance across accuracy, precision, and recall metrics, thus underscoring the potential of skill-inclusive models to redefine credit assessment, promoting fairer financial access and addressing systemic biases in traditional lending.