Optimising Revenue by Efficient Credit Scoring
(2013) FMS820 20131Mathematical Statistics
- Abstract (Swedish)
- The aim of this Master's thesis is to construct an algorithm that nds the optimal credit limit for an individual credit card holder. It should be optimal in the sense that the algorithm suggests the lowest credit limit required for reaching the business goal of generating "X SEK within d days after limit raise" for the issuer of the credit card. This is achieved in three stages. First the probability of default is estimated through a behavioural credit scoring model based on logistic regression. In the second stage the revenue that is generated after a limit raise is estimated using linear regression. In the last stage these estimated quantities
are used for estimating the expected prot generated by the customer, for all
credit limits in... (More) - The aim of this Master's thesis is to construct an algorithm that nds the optimal credit limit for an individual credit card holder. It should be optimal in the sense that the algorithm suggests the lowest credit limit required for reaching the business goal of generating "X SEK within d days after limit raise" for the issuer of the credit card. This is achieved in three stages. First the probability of default is estimated through a behavioural credit scoring model based on logistic regression. In the second stage the revenue that is generated after a limit raise is estimated using linear regression. In the last stage these estimated quantities
are used for estimating the expected prot generated by the customer, for all
credit limits in a xed set of allowed limits. The credit limit that fulls the
business goal is the optimal limit chosen by the algorithm.
The regression models were estimated using real customer data provided by
Resurs Bank, a Swedish niche bank operating in the credit card business. The
resulting performance of the regression models indicate a strong signicance of behavioural variables, leading to the conclusion that the behaviour of a customer can be used for predicting future revenue and probability of default. This information can then be used for nding an optimal credit limit for each customer based on the specic behaviours each individual exhibits. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/3801315
- author
- Skiöld, Henrik and Lundberg, Jimmy
- supervisor
-
- Jimmy Olsson LU
- organization
- course
- FMS820 20131
- year
- 2013
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 3801315
- date added to LUP
- 2013-05-27 15:41:46
- date last changed
- 2013-07-30 10:34:25
@misc{3801315, abstract = {{The aim of this Master's thesis is to construct an algorithm that nds the optimal credit limit for an individual credit card holder. It should be optimal in the sense that the algorithm suggests the lowest credit limit required for reaching the business goal of generating "X SEK within d days after limit raise" for the issuer of the credit card. This is achieved in three stages. First the probability of default is estimated through a behavioural credit scoring model based on logistic regression. In the second stage the revenue that is generated after a limit raise is estimated using linear regression. In the last stage these estimated quantities are used for estimating the expected prot generated by the customer, for all credit limits in a xed set of allowed limits. The credit limit that fulls the business goal is the optimal limit chosen by the algorithm. The regression models were estimated using real customer data provided by Resurs Bank, a Swedish niche bank operating in the credit card business. The resulting performance of the regression models indicate a strong signicance of behavioural variables, leading to the conclusion that the behaviour of a customer can be used for predicting future revenue and probability of default. This information can then be used for nding an optimal credit limit for each customer based on the specic behaviours each individual exhibits.}}, author = {{Skiöld, Henrik and Lundberg, Jimmy}}, language = {{eng}}, note = {{Student Paper}}, title = {{Optimising Revenue by Efficient Credit Scoring}}, year = {{2013}}, }