Loanification - Loan Approval Classification using Machine Learning Algorithms

Banks and other financial corporations have been in the business of lending since the past century. An essential requirement to sustain oneself in such a business is meticulous background checking of a customer before approval of loan failing which the institution lending money carries the risk of non payment of dues by the customer. Such dues which are not repaid back are called bad loans which after 90 days of default become Non Performing Assets (NPA) . According to the Reserve Bank of India ,total NPAs in India are set to reach a twenty year high this year . Thus , the need for developing better models for deciding whether to approve a loan has never been greater . The research work in this paper aims to address this issue and help lending institutions avoid bad loans. A complete background check of each customer has been ensured by feeding entire customer details to our Machine Learning model unlike previous works which relied mostly on specific items like the salary of the client . Ensemble learning models have been the primary focus of this work as Random Forest , Gradient Boosting , XGBoost etc have been used . Out of all the models , the best results were from Gradient Boosting classifier. Finally a Voting classifier has been built by ensembling the aforementioned algorithms . That helped improve the metrics like F1-score and roc- auc-score marginally