With our Predictive Modeling Service, we provide technical guidance on how to solve client decision problems using predictive models. We help clients identify appropriate prediction problems, articulate comprehensive requirements, select appropriate accuracy metrics, design experimental methodologies, validate and prepare data, and advise client technical staff on how best to induce robust predictive models from client data. We also can act as your outsourced predictive modeling staff, to develop predictive models for you.
Customer Success
Using client-supplied historical data, we developed a loan-level residential mortgage default model model with one-twentieth the loss error and more than twice the high-end discrimination of competing models. To satisfy customer requirements, the model’s output probabilities were monotonic in client-designated input variables, such as credit score and loan-to-value ratio.
Using client-supplied historical data, we developed a loan-level consumer bankruptcy model with one-tenth the loss error and three times the high-end discrimination of the customer's extant logistic regression model.