Data Mining in Banking (.ppt)

6,561
-1

Published on

0 Comments
4 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
6,561
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
414
Comments
0
Likes
4
Embeds 0
No embeds

No notes for slide

Data Mining in Banking (.ppt)

  1. 1. Applications of Data Mining in Banking Maria Luisa Barja (Maria.Barja@ubs.com) Jesús Cerquides (Jesus.Cerquides@ubs.com) Ubilab IT Laboratory UBS AG Zurich, Switzerland
  2. 2. Outline <ul><li>Data Mining in Banking </li></ul><ul><li>Application Areas </li></ul><ul><li>Pitfalls in the Development of Data Mining Projects </li></ul><ul><li>An Alternative: A Data Mining Framework </li></ul><ul><li>Open Projects </li></ul><ul><li>Summary </li></ul>
  3. 3. Data Mining in Banking <ul><li>Banks have many and huge databases </li></ul><ul><li>Valuable business information can be extracted from these data stores </li></ul><ul><li>Unfeasible to support analysis and decision making using traditional query languages </li></ul><ul><li>Human analysis breaks down with volume and dimensionality </li></ul><ul><li>Traditional statistical methods do not scale and require significant analysis expertise </li></ul>
  4. 4. Application Areas <ul><li>Four main areas </li></ul><ul><li>Marketing </li></ul><ul><li>Credit Risk </li></ul><ul><li>Operational Risk </li></ul><ul><li>Data Cleansing </li></ul>
  5. 5. Applications: Marketing <ul><li>Objective: </li></ul><ul><li>Improve marketing techniques and target customers </li></ul><ul><li>Traditional applications: </li></ul><ul><li>Customer segmentation </li></ul><ul><li>Identify most likely respondents based on previous campaigns </li></ul><ul><li>Cross selling </li></ul><ul><li>Develop profile of profitable customers for a product </li></ul><ul><li>Predictive life cycle management: </li></ul><ul><li>Develop profile of profitable customers X years ago </li></ul><ul><li>Attrition analysis: </li></ul><ul><li>Alert in case of deviation from normal behaviour </li></ul><ul><li>Techniques : </li></ul>
  6. 6. Applications: Credit Risk <ul><li>Objective: </li></ul><ul><li>Reduce risk in credit portfolio </li></ul><ul><li>Traditional applications: </li></ul><ul><li>Default prediction </li></ul><ul><li>Reduce loan loses by predicting bad loans </li></ul><ul><li>High risk detection </li></ul><ul><li>Tune loan parameters ( e. g. interest rates, fees) in order to maximize profits </li></ul><ul><li>Profile of highly profitable loans </li></ul><ul><li>Understand characteristics of most profitable mortgage loans </li></ul>
  7. 7. Applications: Operational Risk <ul><li>Objective: </li></ul><ul><li>Reduce risk originated by misbehavior </li></ul><ul><li>Traditional applications: </li></ul><ul><li>Credit card fraud detection </li></ul><ul><li>Identify patterns of fraudulent behaviour </li></ul><ul><li>Insider trading </li></ul><ul><li>Detect sophisticated forms of insider trading and market manipulation. </li></ul>
  8. 8. Applications: Data Cleansing <ul><li>Objective: </li></ul><ul><li>Detect outliers, duplicates, missing values,... </li></ul><ul><li>Traditional applications: </li></ul><ul><li>Data quality control </li></ul><ul><li>Detect data values which do not follow the pattern </li></ul><ul><li>Missing values prediction </li></ul><ul><li>Predict values of fields based on previous values </li></ul>
  9. 9. Pitfalls in the Development of Data Mining Projects <ul><li>Data Mining is a process, not a package! </li></ul><ul><li>Expensive, difficult to justify in first instance </li></ul><ul><li>Having substantial parts in common, most data mining projects provide custom solutions that: </li></ul><ul><ul><li>Are more expensive </li></ul></ul><ul><ul><li>Take more time to develop </li></ul></ul><ul><ul><li>Have a higher risk of not being finished </li></ul></ul><ul><li>Ideally, use more than one technique to get a full view of the data </li></ul>
  10. 10. Proposed Alternative <ul><li>Identify the common functionality used for the development of data mining solutions </li></ul><ul><li>Implement and pack this functionality in a way that it can be: </li></ul><ul><ul><li>Reused in many projects. </li></ul></ul><ul><ul><li>Customized to meet the needs of each project. </li></ul></ul><ul><ul><li>Extended, so it grows with its usage. </li></ul></ul>
  11. 11. Object Oriented Frameworks <ul><li>A framework is a reusable, “semi-complete” application that can be specialized to produce custom applications. </li></ul>Framework Ensemble Framework design expertise Programming language expertise OO expertise Domain expertise OO expertise Programming language expertise Framework usage expertise Coding expertise
  12. 12. Data Mining Framework: Benefits <ul><li>Reduces design and development efforts for building concrete applications. </li></ul><ul><li>Lowers threshold for “proof of concept” data mining applications to be developed. </li></ul><ul><li>Allows comparison of results across various methods. </li></ul><ul><li>Facilitates selection of best method(s) for particular domains and business objectives. </li></ul><ul><li>Eases extensibility to new types of methods and algorithms. </li></ul>
  13. 13. Data Mining Framework: General Architecture Project Management Technique implementation Component structure
  14. 14. Data Mining Framework: Component Structure Project Management Technique implementation Data Process Visualization Metadata Component
  15. 15. Data Mining Framework: Method Implementation Project Management Component structure Database Access Data Understanding Data Preparation Modeling Learning Data
  16. 16. Data Mining Framework: Modeling Classification Clustering Regression Prediction Description Learning Data Modeling roles
  17. 17. Data Mining Framework: Open Projects <ul><li>Design and development of: </li></ul><ul><ul><li>A graphical user interface. </li></ul></ul><ul><ul><li>The prediction/description component (based on bayesian networks). </li></ul></ul><ul><ul><li>The clustering component. </li></ul></ul><ul><ul><li>The project management component. </li></ul></ul><ul><ul><li>The preprocessing component. </li></ul></ul>
  18. 18. Summary <ul><li>Data Mining has emerged as an strategic technology for a large bank </li></ul><ul><li>Several business areas where it can be applied </li></ul><ul><li>Application development difficulties </li></ul><ul><li>Proposed a solution based on OO framework technology </li></ul>

×