Underwriting-Loan-Applications

• Developed a machine learning model in R, to assess customer’s financial position and bank’s risk exposure before granting a loan. Built 2 models based on Lasso regression and classification (Random Forest).

• The first model could foresee the percentage of default at an accuracy rate of 35.2% and the second model was able to classify whether the customer will default or not with an accuracy of 72%.

• Finally, tuned these models using an XG Boost to yield an improved accuracy of 62.5% from regression models.