Loan Eligibility Prediction: An Analysis of Feature Relationships and Regional Variations in Urban, Rural, and Semi-Urban Settings

Abstract

In the complex realm of banking and financial systems, the process of extending loans involves navigating through multifaceted criteria that reflect an individual's creditworthiness. As modern global economies evolve, they present distinct financial behaviors across urban, rural, and semi-urban geographies. This dynamic landscape prompts an essential inquiry: Are the determinants for loan approval consistent irrespective of these varying settings? Our research, set against the backdrop of this question, harnesses the power of machine learning to delve deep into this enigma.Utilizing a comprehensive dataset acquired from Kaggle[1], we embark on a rigorous journey of data preprocessing. This phase involves meticulous efforts like rectifying null values and sophisticated encoding techniques to prepare the data for machine learning models. Subsequent stages see the employment of multiple algorithms, including decision trees, random forests, logistic regression, XGBoost, among others, each rigorously trained and critically evaluated for their predictive capabilities. Of these, the Logistic Regression algorithm stood out, boasting an impressive accuracy of 83.78%.However, the essence of this research transcends mere algorithmic performance. The heart of our findings elucidates the subtle yet significant variations in factors dictating loan approvals across urban, rural, and semi-urban datasets. This nuanced understanding underscores a vital recommendation for financial institutions: the imperative to customize and refine their loan eligibility parameters in harmony with these regional intricacies, ensuring a more holistic and informed lending approach.