Predicting borrowers’ chance of defaulting on credit loans
Junjie Liang
(junjie87@stanford.edu)

Abstract
Credit score prediction is of great interests to banks as the outcome of the prediction algorithm is used to determine if borrowers are likely to default on their loans. This in turn affects whether the loan is approved. In this report I describe an approach to performing credit score prediction using random forests. The dataset was provided by www.kaggle.com, as part of a contest “Give me some credit”. My model based on random forests was able to make rather good predictions on the probability of a loan becoming delinquent. I was able to get an AUC score of 0.867262, placing me at position 122 in the contest.

1 Introduction
Banks often rely on credit prediction models to determine whether to approve a loan request. To a bank, a good prediction model is necessary so that the bank can provide as much credit as possible without exceeding a risk threshold. For this project, I took part in a competition hosted by Kaggle where a labelled training dataset of 150,000 anonymous borrowers is provided, and contestants are supposed to label another training set of
100,000 borrowers by assigning probabilities to each borrower on their chance of defaulting on their loans in two years. The list of features given for each borrower is described in Table
1.
Variable Name
SeriousDlqin2yrs

RevolvingUtilizationOfUnsecuredLines
Age
NumberOfTime3059DaysPastDueNotWorse
DebtRatio
MonthlyIncome
NumberOfOpenCreditLinesAndLoans
NumberOfTimes90DaysLate
NumberRealEstateLoansOrLines
NumberOfTime6089DaysPastDueNotWorse
NumberOfDependents

Description
Person experienced 90 days past due delinquency or worse
Total balance on credit cards and personal lines of credit except real estate and no installment debt like car loans divided by the sum of credit