credit scoring is invariably used to answer one key question - what is the probability of default within a fixed period, usually 12 months. It involves assessing an applicant's credit history, payment behavior, outstanding debts, and various financial indicators to determine the likelihood of them repaying a loan. A credit score is assigned to the applicant, reflecting their credit risk and helping lenders make informed decisions about extending credit. Credit scoring can be divided into application scoring and behavior scoring, based on the information used when modeling. Application scoring uses only the information provided in application, while behavior scoring uses both the application information and (past) behavior information.There are three kinds of methods that have been studied in credit scoring; classification techniques, Markov Chain and Survival analysis. Classification has been studied in-depth among these.
Problem Formulation for Predicting Credit Scoring Using Machine Learning: The problem of credit scoring using machine learning aims to develop accurate models that can assess the creditworthiness of loan applicants. It involves addressing challenges such as imbalanced datasets, feature selection, model interpretability, and regulatory compliance. The problem shall be addressed by building a predictive model that can effectively classify applicants into creditworthy and non-creditworthy categories thus minimizing the risk of loan default. During model building, we shall focus on selecting the most relevant features and optimizing model performance.