Bank loan prediction using machine learning

Abstract

The Banking sector is a core foundation of the global economy as it facilitates the financial transactions while providing fundings in various purposes. One key aspect in the banking industry is the ability to accurately predict loan outcomes, which requires assessing the credit worthiness of the loan applicants. Traditional methods of loan prediction are often time consuming, it lacks transparency and interpretability which makes it challenging for the stakeholders to understand the factors that influence loan decisions. With the new addition of machine learning in technology there is an opening to enhance the loan eligibility prediction models and provide transparent insights into the decision-making process. This thesis aims to explore the application of machine learning models and XAI methods for bank loan prediction with a focus on improving accuracy and to get better experience on bank loan applications for both parties. The primary objective here is to develop a robust machine learning model that is capable of accurately predicting loan eligibility and to leverage XAI techniques to explain the reasoning behind these predictions.The research methodology involves a deep analysis from a dataset collected from the internet. The dataset contains various information such as : credit history, loan amount, self employment, earnings etc. Then initial data pre-processing techniques which includes data cleaning or filtering, feature selection and handling values are applied to ensure the quality and the consistency of the dataset. After that, few machine learning models were applied such as: Decision tree, Random forest, NNC etc to build the predictive model. These models are trained and evaluated using appropriate performance metrics such as accuracy, F1-score and AUC(Area under the ROC Curve) score. The goal is to identify the most effective algorithms by comparing them between each other for loan eligibility prediction based on dataset characteristics. Finally, to enhance the transparency and interpretability of the loan prediction models, XAI techniques are applied.These methods facilitate the comprehension of the factors influencing loan decisions, thereby mitigating issues of bias, discrimination, and unfairness. Interpretability techniques, such as analysing feature importance by employing LIME (Local Interpretable Model-agnostic Explanations), are utilised to offer clear and comprehensible explanations for the predictions made by the model. Furthermore, the thesis investigates the ethical implications and fairness considerations associated with loan prediction models. The experimental results demonstrate the efficacy of the proposed approach accurately predicting outcomes while providing interpretable explanations for these predictions. Finally, by utilising machine learning and XAI approaches, this thesis contributes to the subject of bank loan prediction. It provides a complete framework for constructing loan prediction models that are accurate, interpretable, and fair.