Documents

59 views

Financial Services Global Business Unit Analytics and Big Data

Financial Services Global Business Unit Analytics and Big Data. Ambreesh Khanna VP, OFSAA Product Management FSGBU. Program Agenda. Big Data – what does it have to do with OFSAA? Customer Analytics Fraud Default Correlation for Securitized Bond Prices.
of 18

Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Transcript
Financial Services Global Business Unit Analytics and Big DataAmbreesh KhannaVP, OFSAA Product ManagementFSGBUProgram Agenda
  • Big Data – what does it have to do with OFSAA?
  • Customer Analytics
  • Fraud
  • Default Correlation for Securitized Bond Prices
  • Oracle Financial Services Analytical ApplicationsFSDFOFSAA and Big Data
  • Relationship Pricing
  • NBO
  • Reputational Risk
  • Fraud, AML, TC/BC
  • Valuations for Credit Risk
  • Payments Analytics
  • Unified Data Model
  • Use cases
  • OFSAA – Current ArchitectureOFSAAHigh Level ArchitectureUse Case – Customer Attrition2
  • Event
  • Customer gets married
  • Customer Id: 12345Name: Jane DoeMarital Status: SingleOwns house: NNo. of children: 0CASA accountBi-weekly Direct depositAvg. Balance: $10KGold cardLimit: $10KBalance: $7K13
  • Event
  • Customer buys a house
  • Gets mortgage from competing bank
  • 4
  • Event
  • Customer has a baby
  • Opens 529K with competing bank
  • 5
  • Event
  • Customer consolidates accounts
  • Moves all accounts to competing bank
  • Use Case – Customer Retained with Better Insights32Customer Id: 12345Name: Jane DoeMarital Status: SingleOwns house: NNo. of children: 0CASA accountBi-weekly Direct depositAvg. Balance: $10KGold cardLimit: $10KBalance: $7KBank updates customer recordRuns propensity models for NBO and makes time-bound loan offer for $50K for wedding at next point of customer interaction1
  • Event
  • Customer socially announces intent to get married
  • 45
  • Event
  • Customer searches for mortgage on bank website
  • Bank preapproves customer for mortgageMakes offer at next point of customer interaction due to high propensity score6
  • Event
  • Customer announces pregnancy and eventually birth of child
  • 7Bank analyzes purchase pattern and predicts change in status; Augments score with data from social networksMakes 529K offer at next point of customer interaction as per propensity scoreCustomer AttritionFunctional FlowWeblogs, emails, call recordsCore Banking, CRMUser or segment matched Use Case – Trader and Broker Compliance, Internal Fraud1TC/BC/Fraud software monitors patterns of trading activity3Models to find co-relation between events such as large institutional trades and personal calls, or employee accessing a articular customer activity on a regular basis2
  • Additional data points to be provided to TC/BC/Fraud software
  • Emails, SMSs, IMs, weblogs, social updates
  • Use Case – Payments Fraud1Wire Transfer transaction through Bank2Real time fraud detection engine does rule matching and machine learning models try to enhance patterns3
  • Enhanced user profiles and history kept on HDFS
  • Behavior detection models run on Map Reduce
  • Approval/Denial response5Transaction persisted for detailed analytics4
  • Additional data points
  • User, address, geo-location previously known?
  • Any known information from outside the bank about originator or destination?
  • Use Case – Anti Money Laundering1Monetary transactions2Graph analysis is extremely relevant to fraud detectionExtremely large graphs cannot be analyzed with traditional means – order of complexity is likely non-probabilistic in time and spaceSome of these problems are hadoop-able
  • AML software monitors
  • Large cash transactions (CTR)
  • Patterns to identify money laundering (SARs)
  • KYC (checks against negative lists)
  • 4Graph analysis to detect patterns (vertices are entities, edges are transactions)Co-relation between SARs3
  • Additional data points to be provided to AML
  • External information about the customer
  • FraudDiscovery / adhoc Analytical ReportingODBCdTechnical Architecture*M/R“Sqoop”Batch processdHiveQLEndeca / OBIEE*M/R – Map Reduceor EIDEndeca Information DiscoveryFSDF(DB 11.2.0.2+with ORE)Collective-Intellectmove to structured store additional /enriched attributesCIBDA(HDFS/Cloudera ) Hive/NoSQLUnstructured DataBlogsNewsfeedsWatch List ScansFinancial / Marketing /Trade data providers/channelsTrxnsaSource SystemsaIC++ PipesOCI / JNDI-JDBCIbORE native connectivityHiveQLbHiveQLBatchR-connector for Hadoop*M/RNative*M/ROLTPSystemsStochastic Modeling subsystem (with ‘R’ support & ORE connectivity)AAII IbIcAAII IIbScenario Definitions (metadata)SOAPAAIWeb-services interfaces included (WSDL)cBehavior Detection Inline-Processing Engine Post-Processing (pluggable services framework)AAIMSG queuesUsing Big Data to Estimate Default CorrelationPlayers involved in securitization transactions and their rolesRating AgenciesEvaluate credit risk and deal structure, assess third parties, interact with investors, and issues ratingsFinancial GuarantorAsset ManagerInsures tranchesTrades assetsSeniorFundsFundsSPVArrangerMezzaineAssetsPay outsLiabilitiesPay outsFundsPay outsJuniorOriginatorServicerTrusteeBonds with different ratingsCollects & makes paymentsMonitors complianceInvestorsFundsPay outs
  • Prices of Bonds (i.e. tranches) are very sensitive to default correlation of loans
  • We propose to use Big Data comprising of public and private information, Bloomberg and Reuters feeds, commercial transactions, analyst meets, and research reports to estimate default correlation
  • Loans to Textile firmsLoans to Energy firmsLoans to Agricultural firmsEstimating Default Correlation and Securitized Bond Prices – Current StateAnalytical Applications Infrastructure (Data Quality, Metadata, Stress Testing, Modeling, Execution, Big Data Connectors etc)Loans to Energy firmsApplication-specific Processing AreaStaging AreaResults AreaTreasury SystemsDashboards and ReportsCredit Risk EngineCommon Input area for analytical processingLoans to Agricultural firmsValuations EngineCore Banking SystemsDefault MetricsPD, LGD, EAD,Default CorrelationsMarket Risk EngineLoans to Textile firms
  • Company Specific Metrics
  • Demographic, Geographic and Industry information
  • Company Ratings
  • Risky Bond prices floated by firms
  • CDS spreads of the firms
  • Balance Sheet structure and information
  • Basel EngineBond and Tranche Prices,Attachment and Detachment Points,Regulatory ReservesFront Office Systems (like CRM, RTD etc)Stochastic Models to estimate default metricsOBIEEData Quality Checks, GL Reconciliations, Manual Data Adjustments
  • Currently the estimation of default metrics like PD, LGD and Default Correlation only considers structured information
  • Unstructured but rich information contained in Big Data sources like Bloomberg and Reuters feeds and news reports, Analyst comments and Research reports, News on commercial transactions etc. is completely ignored
  • This results in poor default metrics and hence very poor and inaccurate Securitized Bond Prices
  • Securitized Bond Prices are extremely sensitive to Default Correlation, and incorrect estimates of which was one of the main causes of 2008 market crash
  • Estimating Default Correlation and Securitized Bond Prices – Future State Using Big Data SourcesAnalytical Applications Infrastructure (Data Quality, Metadata, Stress Testing, Modeling, Execution, Big Data Connectors etc)Treasury SystemsApplication-specific Processing AreaStaging AreaResults AreaLoans to Energy firmsDashboards and ReportsCredit Risk EngineCommon Input area for analytical processingCore Banking SystemsValuations EngineLoans to Agricultural firmsDefault MetricsPD, LGD, EAD,Default CorrelationsMarket Risk EngineOBIEECompany Specific MetricsFront Office Systems (like CRM, RTD etc)Basel EngineBond and Tranche Prices,Attachment and Detachment Points,Regulatory ReservesLoans to Textile firms
  • Big Data Sources
  • Bloomberg & Reuters feeds and news
  • Analysts comments and Research reports
  • Commercial Transactions
  • Quarterly Investor meets, notes and public announcements
  • Stochastic Models to estimate default metricsData Quality Checks, GL Reconciliations, Manual Data Adjustments
  • Augmenting traditional structured information with the new unstructured information from Big Data sources will result in better estimates of default correlation and PD, LGD, EAD
  • Better estimates of default will result in more accurate prices of Bonds offered to investors via Securitization of assets
  • Estimates of default can be updated quickly as new unstructured information becomes available
  • OFSAA at OpenWorld
  • Monday, September 23
  • 2:30-3:30 Making Sense of the Regulatory Challenges Facing Banks Today & Tomorrow
  • Tuesday, September 24
  • 10:30-11:30 Driving Business Growth by Unlocking Rich Customer Insights
  • 5:15-6:15 Advanced Analytics for Insurance
  • Wednesday, September 25
  • 10:15-11:45 Big Data in Financial Services
  • 4:15-5:15 Use-Case Driven Approach to Using OFS Data Foundation for Data Management Needs
  • Graphic Section Divider
    Advertisement
    Related Search
    We Need Your Support
    Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

    Thanks to everyone for your continued support.

    No, Thanks