Credit Scoring for SAS® Enterprise Miner™
Integrated scorecard development, deployment and monitoring for better decisions
Banking, financial and credit divisions across industries can build, validate and deploy credit risk models using in-house expertise. Scorecard developers and credit scoring managers can make accurate and timely default predictions to: streamline credit approval processes; improve customer acquisition, retention and collections; and reduce exposure to business risk in the organization's consumer lending portfolio.
We have achieved significant economies of scale in terms of resources, personnel cost and time.
Director of Credit Risk and Capital Management Division
Piraeus Bank Group
Comprehensive data preparation.
Save time and resources by accessing, transforming, cleansing and preparing all prerequisite data – including third-party bureau, application, bill-payment and collections data. Data sets, no matter the size, can be examined quickly and easily for patterns, anomalies and missing values through built-in, interactive nodes with many options for exploration, transformation, missing value imputation, outlier analysis and correlation analysis.
Efficient scorecard development.
SAS provides a fast, flexible and economical option to create and deploy credit scorecards for virtually all types of consumer lending products – accounts, cards, loans, mortgages – leading to better credit decisions and reduced losses. You can compute scorecard points for each attribute using either the WOE variables or the group variables that are exported as inputs for the logistic regression model, and you can manually assign scorecard points to attributes.
Variable selection and treatment to quickly understand relationships and behaviors.
Credit Scoring for SAS Enterprise Miner purposely censors the data, making it easy to understand relationships and allowing nonlinear dependencies to be modeled with linear models. This gives the user control over the development process and provides insights into the behavior of risk predictors. The node also screens characteristics so that potentially predictive variables are used while other variables are not.
Automatic creation of the target variable for the rejects data set.
As a necessary step in applying a remedy for selection bias, unrealistic expectations and model overconfidence, SAS offers three industry-accepted ways to infer the rejected data – fuzzy augmentation, parceling and hard cutoff. More robust estimates can be quickly made on how the model performs on both the known population and the entire "through-the-door" population.
Better outcomes and improved portfolio performance.
Assess and control risk within existing consumer portfolios and improve acquisition strategies using SAS advanced predictive analytics techniques. This approach provides a better understanding of the specific risk characteristics and subsequent attributes that lead to delinquency, default and bad debt.