|
Application for Academic
University of Edinburgh
Credit Scoring And Data Mining
Syllabus 2006/2007
Lectures held ; Fridays January 26 ( Thomas), and Feb 2 (Edelman) and 9 ( Thomas)
Aims and objectives
The course aim is to present a comprehensive review of the objectives, methods and practical implementations of credit and behavioural scoring in particular and data mining in general. It involves understanding how large data sets can be used to model customer behaviour and how such data is gathered,stored and interrogated and it use to cluster, segment and score individuals. The aim is to look at the largest application in more detail. Credit scoring is the process of deciding, whether or not to grant or extend a loan. Sophisticated mathematical and statistical models have been developed to assist in such decision problems.
Syllabus
Day 1.
Session 1: Introduction to Data mining and Credit scoring
What is data mining? Databases, data warehousing and data management. Objectives of data mining :Origins of credit and credit lending to consumers; judgmental approaches; introduction of credit scoring; philosophical approach to credit scoring. Overview of use of scoring systems; how credit scoring fits into lenders risk assessment process; what data is needed; role of credit scoring consultancies; testing the scorecard; relation with information system; application form; role of credit bureau; overrides and manual interventions; need for monitoring ; relationship with portfolio of bank products.
Session 2: Statistical Methods for Scorecard Development
Statistical methods in credit scoring and classification methods in data mining; discriminant functions; logistic regression approach; classification trees; non-parametric approaches; graphical models of statistical connections.
Session 3 : Other Credit Scoring Techniques
Mathematical programming and goal programming approaches; neural networks; genetic algorithms and other combinatorial optimisation approaches; expert systems; support vector machines
Lab class on using techniques to build scorecard
Day 2
Session 4: Practical Issues of scorecard performance
Selecting sample; definitions of good and bad; choice of variables; credit bureau data; discarding variables; weights of evidence; coarse classifying continuous variables; taking non-linear functions of variables; dealing with correlations ; reject inference; adjusting cut-off scores; over-rules and their effect on the scorecards.
Session 5 : Measuring Scorecard performance
Hold-out samples, and jack-knifing; bootstrapping; Measuring discrimination- change-over sets; Measuring scorecards –Gini coefficient, ROC curves; Kolmogorov-Smirnov statistic
Session 6: Behavioural Scoring and Profit Scoring
Markov Chain models of repayment and usage behaviour; definition of states of markov chain in repayment behaviour; Orthodox and Bayesian approaches; Mover-stayer and other multi-class models. Profit scoring, generic scorecards; including economic information in scorecards; #p#分页标题#e#
Lab class on coarse classifying and variable choice
Day 3
Session 7:Survival analysis approaches and customer lifetime values
When not if events occur; survival analysis; proportional hazards models, use in profit scoring; application to customer lifetime value
Session 8: Basel Accord and other applications of scoring methodology
Credit risk modelling; Basel Accord and impact on credit scoring .Debt recovery; credit extension; fraud prevention; provisioning for bad debt; transaction authorization; pre-approval; mortgage scoring; small business scoring; credit reference guarantees.
Direct marketing; prisoner release; housing allocation; university admissions. proteomics
Session 9: General Data mining objectives and algorithms
Task, structure, score function, optimization methods, data management techniques. Clustering, regression, classification and data. Basket analysis, share of wallet. Non credit scoring examples of data mining in business
Indicative reading list
L.C.Thomas, J.N.Crook, D.B.Edelman, Credit Scoring and its Applications, SIAM Press, Philadelphia, (2002)
H.McNab, A Wynn, Principles and Practice of Consumer Credit Risk Management, CIB Publishing, Canterbury (2000)
S.Jacka, D.J.Hand, Statistics in Finance, Edward Arnold ,1997.
L.C.Thomas, J.N.Crook, D.B.Edelman, Readings in Credit Scoring , OUP, (2004)
E.M.Lewis An Introduction to Credit Scoring, Athena Press, San Rafael, (1992)
E. Mays. Credit Scoring for Risk managers, South Western, Mason, (2004)
D.M.Hand, H.Mannila, P.Smyth, Principles of Data mining, MIT Press ( 2001)
Student Learning Outcomes
Understanding of basics of data mining
Knowledge of real application of data mining, including clustering, segmentation and scoring.
Understanding of statistical and alternative methods of constructing scoring rules.
Understanding how to process data prior to model building.
Ability to assess and monitor a scorecard.
Awareness of current and new applications of credit scoring techniques.
Summary of teaching and learning methods
The unit is delivered through pre-course reading, personal reflection, lectures including group exercises discussions and case studies, and computer laboratory class.
Summary of assessment methods; 100% Project
Project will be handed out at start of session on February 9 2007.web：http://www.ukassignment.org/daixieAssignment/daixieyingguoassignment/
|
