ahananamburi
commented
6 years ago Incorporated the feedback by narrowing down on the problem. Addressed dynamic pricing and credit limit changes based on risk as a specific solution.
Closed TomHunter1 closed 6 years ago
ahananamburi
commented
6 years ago Incorporated the feedback by narrowing down on the problem. Addressed dynamic pricing and credit limit changes based on risk as a specific solution.
I agree the financial industry could certainly benefit from the applications of Machine Learning. There's lots of use cases posted to kaggle or used by udacity as projects (Some examples below). You can always scour them for new ideas so you have a rich body of work to look at.
Bank of America / an old subgroup of MBNA, use logistic regression models to do exactly what you're talking about and these feed into the greater system of marketing decisions for direct mailer etc., but they've been doing that since the mid-90's. My challenge is how do you define the problem to be more specific. ML on credit lending data is tried and true, there are lots of open source datasets and then Equifax, Experian and Transunion would be good data sources and they do sell credit histories that are run though ML models to determine the likelihood of granting credit access. Could you narrow in on a more specific problem? Maybe incorporating this idea with other ideas like a social issue? Predatory high interest loans is a major issue in low income neighborhoods.... could you segment out people with high interest loans, give them a micro loan with a reduced interest rate to allow them to get out of debt. These people are obviously high risk, could you use ML to predict who has the highest probability to succeed?
Approaches of loan classification open source datasets / articles: https://www.kaggle.com/c/loan-default-prediction https://www.kaggle.com/wendykan/lending-club-loan-data https://nycdatascience.com/blog/student-works/kaggle-predict-consumer-credit-default/