SAS Credit Scoring for Banking

SAS Credit Scoring for Banking is the solution of a problem of credit scoring for banks including functionality on maintenance of all necessary processes. The solution includes means of processing and information storage, to formation of data marts, a broad set of analytical tools for creation and the analysis of models of credit scoring and extensive reporting system for solving of tasks of assessment of operability of scoring models and a status of the loan portfolio.

Description

The SAS Credit Scoring for Banking tools allow the specialist to create in the field of intelligent data analysis models of credit scoring for consumer loans, credit cards, overdrafts, automobile, mortgage and other credit products. Scoring is directed to the solution of different tasks of assessment of probability of a default of the client before determination of strategy of work of collection division and creation of a rating system according to the recommendations of Basel committee.

The main commonly accepted models are divided into the following types, each of which can be implemented by means of SAS Credit Scoring for Banking:

• Biographical (application) scoring
• Behavioural scoring
• Collection scoring
• Anti-fraudulent scoring

The PD, LGD, CCF/EAD models according to Advanced-IRB approach of Basel committee Data are the cornerstone of any predictive model, and the accuracy of the received forecasts directly depends on their quality. This simple rule often is a key to creation of stable and reliable model. To solve possible problems of quality of data and also as much as possible to reduce costs for their collecting and consolidation, SAS Credit Scoring for Banking offers the users a logical structure of data – DDS. All necessary information for DDS arrives from operating systems, the application systems and other bases of the client through the special ETL procedures of collecting, processing and loading.

DDS represents the uniform source of the consolidated information organized in the form of a logical structure of data. Such organization allows to use DDS as a reliable basis for creation of solutions of SAS. In particular, it belongs to formation of data marts for the solution SAS Credit Scoring for Banking. Thanks to the predeterminated ETL procedures, the user the user-friendly graphical interface of creation of selection is available to generation of variables to modeling, to a research with in advance prepared list the most often used from them.

For data analysis and development of the SAS Credit Scoring for Banking models offers the users idle time in use, but at the same time very flexible and multifunction tool - SAS Enterprise Miner. SAS Enterprise Miner has the intuitive graphical interface for creation of projects on data mining and modeling.

During the work with SAS Enterprise Miner the analyst creates the charts consisting of data sources, nodes of processing and indications of the direction of the movement of a data stream. Nodes of processing represent ready-made solutions of separate subtasks of the analyst with a possibility of setup of parameters and the choice of algorithms. All nodes are broken into the groups making the logical sequence of development stages of model – SEMMA:

In addition to the standard SEMMA nodes, SAS Credit Scoring for Banking include special nodes for development of scoring cards:

Interactive Grouping node

Selection of variables on the basis of calculation the statistician of Information Value or Gini at choice of the analyst Splitting variables into equal groups (Fine Classing) Optimal consolidation of groups and formation of final classes (Coarse Classing) by means of decision trees Calculation of Information Value and Weight Of Evidence for groups of variables Possibility of interactive change of Fine Classing and Coarse Classing, manual editing of splitting with instant recalculation of all the statistician

Scorecard node

Creation of the scoring map with use of logistic regression Possibility of the parametrized scaling (calibration) of the received scoring card Detailed statistics and diagrams with model quality evaluations (Gini, AUROC, KS, etc.), a possibility of the choice of point of cutting off

Reject Inference node

Possibility of carrying out Reject Inference one of 3 methods at choice of the analyst A broad set of the preset reports of validation which are automatically updated on a regular basis is available to users of SAS Credit Scoring for Banking. For simplification of work with reports, all key indicators have the configured color indicators showing this or that test is how well passed. In spite of the fact that the preset reports already contain full information on a model status, the user has a possibility of adding of reports and calculation own the statistician.

SAS Credit Scoring for Banking the applicant champion for comparison of models on any reporting date allows to use a technique. Comparison is carried out as using overall assessment of work of model, and separately on each indicator for the selected date.

Each report contains values corresponding the statistician, color indicators and the configured set of diagrams, displaying both static parameter values, and dynamics of their change. Besides, SAS Credit Scoring for Banking allows to carry out validation not only on all portfolio, but also on any its subsegment by means of a task of business filters. Thereby the analyst has an opportunity not only to reveal strong and weaknesses of model, but also to study those segments of clients for which it needs adjustment.