Credit Risk Default Prediction

Inspiration

Americans total credit card balance is $986 billion in Q4 of 2022 Average American has $2948 credit card debt The delinquency rate for Q4 2022 is 2.25% So it’s a whopping $22.17 billion in a single quarter!! Our machine learning model helps for companies to perform better and lessen the dent in economy.

What it does

To improve our profitability and mitigate credit risk of companies through the power of predictive systems.

How we built it

We built it using American express dataset, which contains 5.5 million records with 458k unique customer's 12 month data, aggregated features by median and ran both xgboost and neural network models to predict the probability of a customer

Challenges we ran into

The dataset was pretty huge, so we loaded into chunks, and the training time for both the algorithms was pretty high.

Accomplishments that we're proud of

The model accurately predicts the probability of a customer defaulting by 92% and we have significantly reduced the training times of the model by 40x

What we learned

That data processing and cleaning is pretty hectic task. Also for every machine learning solution there needs to be a business strategy associated with it. We are proud to create a business strategy for a company which improves their revenue and reduces the default rate.