SAS Data Warehouse for Risk analysis in Vilniaus Bankas AB
Transcription
SAS Data Warehouse for Risk analysis in Vilniaus Bankas AB
SAS Data Warehouse for Risk analysis in Vilniaus Bankas AB Vilniaus bankas www.vb.lt Agenda 1. A little bit about our bank 2. Situation in Lithuania banking sector 3. Risk management concepts as we take in our project 4. Architecture overview 5. Most important part of project Data Model 6. SAS tools in use (some examples) 7. Data mining+DW administrator + IntrNet=excellent solution to our bank 1. Largest bank in Lithuania 2. Have activity in all banking sector 3. Problems transforming data to information as usual Some results for 1999 Profit before taxes increased by 34.7% to LTL 80.5 million Net profit up by 24% to LTL 74.2 million Assets of the Bank increased by 21% Group's assets increased by 77% to LTL 5 billion during 1999 Earnings per Share LTL 4.95 Bank's Return on Equity 16.02%, Return on Assets 2.53% Loan quality remains high despite the turbulent markets Strong contribution from Bank's subsidiaries Merger with Bankas Hermis completed Very popular question. Data is everywhere but where is information ? Vilniaus Bankas Branches and the transactional system is distributed Banking sector more and more complicated Lithuania banking sector. What situation we have? Market consolidation (so our portfolio is growing and becoming more complicated The new products and services emerge (portfolio is becoming more complicated) The entrance to market new players(competition increasing ) The volumes of data that we have on the hand is growing New transactional systems emerge. So how to consolidate them ? Decision support systems only in plans We want the system for decision support What choices we have? To buy new system ? To make everything with our programmers and technology we know ? Make a contract with third party company ? Any other solution? - Expensive - The specialises system is difficult to integrate concerning banking management methodology (usually the is no technical difficulties) - Expensive maintenance and new subjects creation and integration + The is no headache for IT + We have starting methodology know-how on hand +Guaranteed maintenance To buy new system? To make everything with our efforts? - know-how shortage - long project maturity - The is headache for IT + Guaranteed maintenance + cheap maintenance + Good possibility integrating with existent systems Our choice Do yourself Business process analysis Primary sources analyses Design of data model Methodology Administration SAS representatives SAS installation Programming Teaching !!! Shareholder value added That is our approach to project Deep analysis of reporting processes in bank helped as to make efficient reengineering of reporting and after that efficient cost savings We tried to minimise our need for application customisation and maximise usage of SAS EIS • More efforts to teach users how to work with standard SAS tools helped us to minimise need for programming • After standard reports goes by the push of the key we started to concentrate on risk management system development Risk factors that we want to manage • • • • • • 28.04.00 Interest rate Market Liquidity Credit Operational Capital But the first thing we must to think it is quality of data So how we done it ? • First we decide to move existing finical reports from transactional system to warehouse environment • But the data models of warehouse was not transactional system but Risk control in mind Architecture overview Financial analysis and risk management mart Profit / lost and internal transfer model Portfolio scenarios modulation Interest rates analysis FX analysis Index Time series Reporting modulation generator models DW Multidimensional portfolio analysis Liquidity analysis Cash flow analysis Tasks Build in Multidimensional Objects (methods and data) reports analysis Additional error check DW CAPTURE TRANSFORM CLEAN BUILD Bank transactional system Investment banking Reuters KONDOR+ Cards mang.system PRIME Leasing company Life Insurance Primary sources - DW The foundation to datamodel Architecture Business Concept Analyze (project) Design (project) A B Conceptiualmodell C C’ D Logical datamodel Physical datamodel A-level - Nine data concepts A B C C’ D Involved Involved Party Party(IP) (IP) Arrangement Arrangement (AR) (AR) Location Location (LO) (LO) Condition Condition (CD) (CD) Classification Classification (CL) (CL) Event Event (EV) (EV) Product Product (PD) (PD) Business BusinessDirection Direction Item Item(BDI) (BDI) Resource Resource Item Item(RI) (RI) Involved Party Arrangement Condition Product LTL % Location Classification Event SEB Low Interest Home Loan Business Direction Item Resource Item One SEB Share B-level DM - Three hierarchy's Involved Party Main Descriptor Occurrences within one data concept Additional information about occurrences within one data concept Relationship Relations between occurrences within or between data concepts A B C B-level in -DM - Main-hierarchy - C’ INDIVIDUAL EMPLOYMENT STATUS TYPE D Individual INDIVIDUAL MARITAL STATUS TYPE Organisation INVOLVED PARTY TYPE Organisational Unit Employment Position Working Individual Not Employed Individual Married Individual Unmarried Individual B-level-scoping - Map the user needs to DM - m1 DM Modelled data concept in the risk mart-case Involved Involved Party Party(IP) (IP) Arrangement Arrangement (AR) (AR) Location Location (LO) (LO) Condition Condition (CD) (CD) Classification Classification (CL) (CL) Event Event (EV) (EV) Only a subset of the data concepts are being modelled. Product Product (PD) (PD) Business BusinessDirection Direction Item (BDI) Item (BDI) Resource Resource Item Item(RI) (RI) C’-level in DW - Business Objects - A B C C’ D Involved Party Arrangement Condition Product Individual Product Arrangement Resource Item Accounting Object Transaction Campaign Segment Address A B Map conceptions in DM against DW in m1 C m1 C’ D SEB-DM Begrepp SEBDW Logisk C’-modell Analysis tools we used • User takes date from DW with SQL filters and reuse tools we build – SAS AF – SAS EIS – SAS BASE – SAS GRAPH – SAS STAT – SAS EIS • The main tool SAS EIS Some examples • The situation was extreme So we build first Cards and Merchants online monitoring system • Risk management system Cards and merchants management Operational risk management Good performance,online access mode , interacting with online s as little as possible,there is no what report we need,from INFORM possibility to learn,flexibility, statistics Alternatives: INFORMIX,PROGRES, JavaScript, Access, VB , C++ SAS EIS,AF Some opinion + SAS EIS,DWA project time as short as possible + ETS, STAT, Data mining from simple to complex risk management models realisation + Start in PC , and after then go to more and more powerful server platform + Good maintenance (my practical experience) + Everything from one company Main tasks we have • Take data from INFORMIX (UNIX) data bases without disturbing authorization system • Show only changes from last snapshot • Multidimensional analysis • Maintenance of archive • Take analysis from archive with the same tools • The system must be able to learn from cheat facts • Ho the system learns – statistics is calculated – we are setting priory identifications rules – with a new data statistics is recalculated – when we find the crime, system changes weight to statistics • The main tool EIS Risk management mart Most important concerns You must to have leader who understands • Business • Risk • System analysis • Design • and have CEO’s support You must to have team of specialists • Business • Risk • System analysis • Design ! Otherwise you will have substitute Risk management mart The first question from business Can we find there some hidden money? The answer Wrong question ! 28.04.00 Dimensions that we use in measure risk • • • • Lost profit Shares price on the market Economical share price Others financial indicators as ROE, ROA 28.04.00 VaR (Value at Risk ) ? • The question is very simple • What is our portfolio value at fixed maturity with some fixed probability measure – Answer profit loss empirical probability probability function – J.P. Morgan/Reuters RiskMetrics methodology as most know to business people – For most analytically thinking users SAS have tools to go more deeper 28.04.00 If we look at basic VaR methodology blocks ? The horizons of our prognoses and confident intervals Transformation of financial portfolio instruments to cash flow and some index association Transformation of cash flow to some fixed time intervals History of financial indexes 28.04.00 Evaluation of portfolio variation VaR (reports) Results about financial instruments variations and correlation ! VaR VRV P T V ! "( p1 w1$ 1m ),!, ( pn wn$ nm ) * 1 ( ! R !( ( , 1 , n +1 ( ) , 1, n rp ! rp ! 28.04.00 , n +1,1 ! ! ! 1 , n +1 , n , n ,1 ' ! %% ! % % 1 & n - (1 . / i !1 n - (1 . / i !1 ! ! i )ri n i 0 )ri . 0 . 5 2 - 3i 2 Pi 2 ri 2 i !1 1 # Data Warehouse from risk management point of view • Consolidated reporting without programming • Derivatives portfolio management • Good performance processing huge amount of data • VaR • Monte Carlo Data mining ? • Huge amount of data analyses • Statistics + Neural networks SAS IntrNet • • • • • Administration minimum SAS interface In internet No need for Java coding We are using Netscape WWW server Standard tool remains in use