Ultimate Scorecard Development Tool - Scorto Credit Decision - Now Optimized For Insufficient Data
Scorto Credit Decision software for in-house credit scorecard development has been optimized for working with insufficient data.
By: Scorto Corp.
In today’s post-crisis period, when existing credit scoring methodologies are not applicable and standard approaches are no longer efficient, an accurate scoring model, optimized for the organization’
Scorto Credit Decision v.3.2 (http://www.scorto.com/
“These days, when there is still very few data collected within the after-crisis period, it is more than ever important to get the correct scorecards having a small amount of data available,” says Andrew Antonets, Credit Decision Product Manager.
The latest Credit Decision release implements the professional approach to transform categorical variables into numeric using the Weight of Evidence (WOE) measure and to control WOE correlation in modeling. This approach was by now available only through a narrow range of specific top-expert tools or data mining software. From now on Credit Decision enhanced with this feature provides significant increase in informative value of input data, therefore an increase in amount of parameters used by the model.
The new version also allows obtaining additional information on model stability and adequacy:
- Confidence intervals for analyzed events in building the strategic curve allow for better assessment of model stability and adequacy on the relevant score ranges. Nowadays, when the strategy curves are never as smooth as it was seen before and confidence intervals can drastically differ for different scorebands, it is more than ever important to understand whether this fits to confidence criteria or not.
- Marginal Chi-Squared reports for categorical characteristics allow controlling model stability on the key portfolio subsets. In the post-crisis times, when Gini values for the current scorecards in production often deteriorate, this report is the best guide to define whether the model performance can be improved to the “before-crisis”
- Another option for Probability of Default (PD) estimation. PD estimation with the logistic function extends the scoring range calibration functionality, allowing, in some cases, for more correct PD estimations.
Another improvement made is targeted to ensuring the utmost effective columns transformation for the insufficient data processing.
“These days scorecard developers need to squeeze maximum from the scarce data they have for the post-crisis period, that is why more versatile columns transformations are required,” comments Mr. Antonets. “The new version of Scorto Credit Decision grants this opportunity with extended numeric field transformation functionality, including advanced conditional function ‘if’ and a tool for special numeric characteristics calculation.”
To facilitate creating a large number of scorecards and reports on a dataset, from now on Scorto Credit Decision allows setting default processed dataset and model parameters.
And the last but not the least announcement is adding MySQL DBMS to the list of supported databases, especially valuable nowadays, when many financial institutions and even departments in large banks are forced to move to open-source databases due to tighter budgets.
More details and a free consultation on creating in-house scorecards with Scorto Credit Decision can be requested at Scorto Corp. website - http://www.scorto.com
# # #
About Scorto Corporation
The Las Vegas, NV-based Scorto Corp. is an innovative provider of complete decision management and credit scoring solutions for the banking, consumer lending, and retail industries. Being a direct descendant of Alyuda Research (http://www.alyuda.com), one of the world leaders in the development of neural network forecasting and data mining software, Scorto actively uses its huge scientific potential and expertise. Today Scorto spreads its activity globally serving customers from 4 continents, with its offices in the USA, Europe and Asia and is well-known for its prominent R&D expertise.