Начало / Университетът / Факултети / Стопански факултет / Архив / Архив "Стажове и стипендии" / Credit Scoring Summer Practice: Experian Bulgaria – Regional Analytics Group





Credit Scoring Summer Practice

Experian Bulgaria – Regional Analytics Group


For 2nd/3rd year Bachelor’s and Master’s students in Mathematics, Econometrics, Statistics, Finance or related.


Practice duration: 2 weeks in July (20.07.2015 – 31.07.2015, 3 sessions per week)

Total classes: 24 astronomical hours (4 hours per session)

Venue: Dedicated Computer Lab at the Office of Experian Bulgaria


The Practice has been designed to mirror a typical case of an Application Scorecard Development Project. As such, it is most suitable for students interested in Statistics, Econometrics, Probability theory and Applications, Credit Risk Modelling, and Mathematics for Business Applications. The aim of the Credit Scoring Summer Practice is for the participants to gain an understanding of the standard Scorecard Development process at Experian, as well as familiarity with the internally developed software Modelling Development Studio as an analytical tool. Besides the Scorecard Development methodology and tools, students would also get exposure to the best practices followed throughout the development process, which ensure that a Credit Scoring solution is congruent with Business logic. Each topic is presented by Analysts of the Regional Analytical Group of Experian, and this gives further insight to students on the typical tasks, issues, and decision making process that a standard Scorecard Development process involves.


The practice comprises in total of 6 demo sessions, each one of 4 hours, which take place at a dedicated computer lab at the Experian Office in Sofia. The covered topics are as follow:


Week 1:

  • Session 1: Overview of Credit Scoring and Project specification

Introduction to Credit Risk Modelling, Scorecards Development, Implementation, and Monitoring. A brief overview of different types of financial risk is also presented. Finally, students are acquainted with the goals of the Business Case Study, and the Project Specifications, which detail out the scope of the tasks to be performed throughout the Practice.


  • Session 2: Introduction to Modelling Development Studio and Data set up tasks

The session focuses on setting up the building blocks for work with the Modelling Development Studio. It presents an introduction to the MDS software and its functionalities, and guides the students throughout the preliminary steps of Data import and MDS System set up. Additional tasks comprise of importing already prepared objects for preliminary data investigation, such as Fine and Coarse Classes of modelling variables.


  • Session 3: Initial data analyses tasks and Introduction to Modelling

Students would be guided to create by themselves Fine and Coarse Classes on the pre-selected set of variables. An overview of the standard Credit Scoring Methodology used at Experian would also be presented.


Week 2:

  • Session 4: Known Good/Bad and Accept/Reject Models Development and Assessment

The session focuses on the basic building blocks of a Credit Scoring Model developed at Experian. These are the so called Known Good/Bad model (assessing the probability of default on the population with known payment history), and the Accept/Reject Model (estimating the probability of a potential credit applicant to be accepted). The corresponding score coefficients are estimated through Stepwise Linear Regression technique, and the respective robustness checks (significance level, over-fitting measures) would also be discussed.


  • Session 5: Reject Inference overview and Final Model creation

Once the Known Good/Bad and Accept/Reject Models are developed, the Practice continues with an introduction to the process of Reject Inference, that is – the process of defining an inferred probability of good (non-default) for all rejects. The final scorecard model (also known as Parcelled model) is then developed on accepted and rejected applicants using the known (accepts) and inferred (rejects) probability of good.


  • Session 6: Final model Validation Reports, Cut-off strategies, Final Documentation

Once the Final Scoring Model is developed and assessed – further validation checks are performed in order to ensure its robustness. It is checked whether the model would produce similar results if applied on a different sample. The session also presents the use of Cut-off reports in MDS for comparison between existing decision making process, and the proposed Scorecard. Finally, an example of Final Documentation and set of reports is presented.


If you are interested to apply, please send short resume and motivation letter to Dimitar.Vasilev@experian.com by 25th of May 2015. All applicants will receive feedback by 15th of June.