Debt Repayment

Model prepayment risk and automate document classification to predict repayments/prepayment risks of high interest debt (eg, credit debt) and fixed-income assets (eg, mortgage backed securities)

Predict whether or not someone will apply for a mortgage over given time periods

Score debt to reduce repayment and default risk

Achieved predictive accuracy of repayment model over 76%

Reduced model deployment time by 70%

H2O.ai identified $10M of debt retired for single customer

Reduced risk of mortgage prepayment due to the loss of future interest payments and the creation of complications in the refinancing strategies. 

Artificial neural network models, which are applied to predict the prepayment rate on pools of mortgage loans. 

Logistic regression for predicting categorical outcomes, using a number of explanatory variables to predict the values of the target variable that can be either binomial or multinomial.

Model interpretability by conducting a sensitivity analysis to determine how input variables affect the output of the model.