Transcription
Susara van den Heever, PhD6 October 2015Decision Optimization: the Key Differentiatorfor Next-Generation Analytics 2015 IBM Corporation
Important DisclaimerIBM’s statements regarding its plans, directions, and intent are subject to change orwithdrawal without notice at IBM’s sole discretion.Information regarding potential future products is intended to outline our generalproduct direction and it should not be relied on in making a purchasing decision.The information mentioned regarding potential future products is not a commitment,promise, or legal obligation to deliver any material, code or functionality. Informationabout potential future products may not be incorporated into any contract. Thedevelopment, release, and timing of any future features or functionality describedfor our products remains at our sole discretion.Features under development forpossible future release; confidentialuntil announced2CONFIDENTIAL“unmarked”e.g. Statements of strategy,competitive standing, etc.e.g. promotional material andcontent you will find in other publicdocuments 2015 IBM Corporation
Overview Introduction to IBM Advanced Analytics How Decision Optimization on Cloud is changing the Prescriptive Analyticslandscape The importance of uncertainty: the Uncertainty Toolkit Cognitive insights: a vision for consumable Watson Analytics-like optimization3 2015 IBM Corporation
Overview Introduction to IBM Advanced Analytics How Decision Optimization on Cloud is changing the Prescriptive Analyticslandscape The importance of uncertainty: the Uncertainty Toolkit Cognitive insights: a vision for consumable Watson Analytics-like optimization4 2015 IBM Corporation
IBM Advanced AnalyticsThe most comprehensive portfolio of data and analytics offeringsAnalyticsPlatformDescriptive,Predictive, &PrescriptiveAnalyticsSoftware &ServicesOpenSourceInitiativesAnalytics /IndustrySolutionsPre-built solutionsbased on AnalyticsPlatform, e.g.PredictiveMaintenance &Quality (PMQ),PredictiveCustomerIntelligence (PCI),Energy Platform,and moreWatsonCognitive baseInternet ofThingsA platform designedfor data-richsolutions that areAbility to learn,enhance, scale and secure andaccelerate, human intelligentexpertiseInsightCloudServicesIndustry apps andsolutions built inpartnership withThe WeatherCompany andTwitter thatdeliver newinsightsCloud Data ServicesCloud-based experience that enables developers to use the right platform-as-a-service tools formobile and web app development5 2015 IBM Corporation
Unified Advanced Analytics PlatformDifferentiated Analytic Solutions Made Possible By Breadth, Depth and ScalePredictiveMaintenance &Quality cations mSolutionsDifferentiated Analytic ration and Deployment Services6 2015 IBM Corporation
ALL Data(Structured,Unstructured,Streaming)IBM Analytics Platform:End-to-End Analytic CapabilitiesDataPreparationAnalyticsat Scale Predictive Data connectors models Data Wrangling StatisticalAnalysis Data models DecisionOptimization Analytic models7Insight toActionALL l-Time) Real-TimeScoring OptimizedDecisions APIs & services Dashboards /Interactive apps 2015 IBM Corporation
Business Analytics and OptimizationStochastic OptimizationPrescriptiveCompetitive AdvantageOptimizationPredictive ModelingPredictiveForecasting/ ExtrapolationStatistical AnalysisAlertsQuery/ Drill DownDescriptiveAd hoc reportsStandard ReportComplexity88From the book “Competing on Analytics” by Thomas H. Davenport , Jeanne G. Harris 2015 IBM Corporation
IBM’s Decision Optimization OfferingCPLEX OptimizationStudio (COS)Optimization EngineDecision OptimizationCenter (DOC)Development &Deployment PlatformModel complex business problems.Solve with IBM CPLEX Optimizer.Prescribe precise and logicaldecisions.Build optimization solutions.Includes data analysis & visualization,scenario management, collaborativeplanning, and what-if analysis.Decision Optimization onCloudSaaS DeliveryAdvanced Tools forOptimization underUncertaintyPrescriptive analytics as a service.No install, no setup.Embed in other applications.9Compare multiple plans, scenarios,KPIs.Understand trade-offs e.g. costs vsrobustness.Improve solutionstability/robustness. 2015 IBM Corporation
Overview Introduction to IBM Advanced Analytics How Decision Optimization on Cloud is changing the Prescriptive Analyticslandscape The importance of uncertainty: the Uncertainty Toolkit Cognitive insights: a vision for consumable Watson Analytics-like optimization10 2015 IBM Corporation
IBM Decision Optimization on Cloud (DOcloud)(30-day free trial https://ibm.co/docloudtrial)Flexible configuration:virtual machineor bare metal servers,# cores and RAM optionsFlexible pricing:pay-as-you go,committed hours,reserved capacityDropSolveDOcloud APIDOcloud CommunityDrag-and-drop tosolve models oncloudEmbed DOcloud inany applicationQuick access to OR, IT,and cloud experts11 2015 IBM Corporation
How does DOcloud fit with COS and DOC?DOcloudCOSDOCCustom DOcloud solvepossible via DOcloudAPIDOcloud solve fromCPLEX StudioModeling for DOcloudsubscribersCOS Community Edition freedownload for limited size12IBM Confidential 2015 IBM Corporation
DOcloud personasI want to build models &solve them faster, withmore capacityOscarOR ExpertI want to integrateOscar’s model into theend-to-end solutionDeniseDeveloperI want to understand howthis impacts the bottom line,and save on softwarelicense costsLiamLine of Business13I want to create anoptimization-basedapplication and sell itEthanAnalyticsEntrepreneur 2015 IBM Corporation
Typical pain pointsI need moresolve capacityOscarOR ExpertWhy can’t Oscar set thiscloud server up himself?I’m swamped.DeniseDeveloperWe only solve this hugeproblem once a year – I cannotinvest in a server just for that.LiamLine of Business14I want to provide a CPLEXbased application to mycustomers, without requiringthem to buy a license.EthanAnalyticsEntrepreneur 2015 IBM Corporation
The six user experiencesDiscover,Try, andBuySupportGetstartedLeverage &ProductiveExtendUse(API)Manage&UpgradeHow does Decision Optimization on Cloudchange the user experience?15 2015 IBM Corporation
How does DOcloud change the user experience?Discover,Try, andBuySupportGetstartedLeverage e: buy online with a credit cardOr if you prefer, you can still talk to an IBM representative as beforeFree trial: https://ibm.biz/trydocloudNo installation or download required!16 2015 IBM Corporation
How does DOcloud change the user experience?Discover,Try, andBuySupportGetstartedLeverage &ProductiveExtendUse(API)Manage&UpgradeDropSolve: easy drag-and-drop interface to solve models on cloudOnline documentation & samples17 2015 IBM Corporation
How does DOcloud change the user experience?Discover,Try, andBuySupportGetstartedLeverage &ProductiveExtendUse(API)Manage&UpgradeMP and CP on cloud, with choice of VM or bare metal servers on IBM SoftLayerPay for what you need – scale up or down according to demandInfrequent-use excess capacity available on cloud, e.g. annual planning: no investment in additionalonsite capacity required18Reduced total cost of ownership! 2015 IBM Corporation
How does DOcloud change the user experience?Discover,Try, andBuySupportGetstartedLeverage &ProductiveExtendUse(API)Manage&UpgradeNo installation, maintenance, upgrades – all managed by IBMContinuous delivery means the latest version is always on the cloud!19 2015 IBM Corporation
How does DOcloud change the user experience?Discover,Try, andBuySupportGetstartedLeverage &ProductiveExtendUse(API)Manage&UpgradeDOcloud API: embed CPLEX Optimizer on cloud in any application (on-prem or cloud)Option to complement with other IBM Analytics services on cloud (e.g. SPSS scoring, BI)Option to deploy optimization-based services/applications which embed DOcloud, on IBM Bluemix20 2015 IBM Corporation
How does DOcloud change the user experience?Discover,Try, andBuySupportGetstartedLeverage &ProductiveExtendUse(API)Manage&UpgradeSame support channels as beforeAccess to wide range of OR, IT, and cloud experts within IBMdeveloperWorks community: https://developer.ibm.com/docloud/21 2015 IBM Corporation
Summary: Prescriptive Analytics on the Cloud Reduced IT spend No license feesNo maintenance, upgradesReduced on-prem hardware spendReduced IT expertise to set up / maintain server environmentPay-per-use options Increased consumability Easy access (no installation / setup)Flexible embedding via Rest APIsEmphasis on communities / collaborationCombine with other analytics and data services for end-to-end applicationsOptimization combined with cloud & user-centric analytics platforms open upa new world to LoB, data scientists, OR & IT experts22 2015 IBM Corporation
Overview Introduction to IBM Advanced Analytics How Decision Optimization on Cloud is changing the Prescriptive Analyticslandscape The importance of uncertainty: the Uncertainty Toolkit Cognitive insights: a vision for consumable Watson Analytics-like optimization23 2015 IBM Corporation
Uncertainty affects all industries Plans are out of date assoon as they are created.No matter how preparedwe are, each planningcycle is afflicted by futureuncertainties - prices,demand, supply, weathereffects We don’t have a clearvision of possible futurescenarios, and their effecton plans.We get caught in a costly reactive cycle where we fix issues after the fact,instead of anticipating and planning for them.24 2015 IBM Corporation
The Uncertainty Toolkit Create plans to anticipate the future Hedge against uncertainty in future scenarios Reduce reactive changes Increase margins Reduce response time Compare and evaluate alternative plans based on KPIs Risk-reward trade-offs across future scenarios25 2015 IBM Corporation
Uncertainty Toolkit origins 2013/14 Joint Program between IBM Research and Decision Optimization Goals Increase customer solution resilience, reliability, and stability Improve trust & understanding of optimization technology Approach Leverage Decision Optimization & mathematical optimization to hedge againstuncertainty (e.g. uncertain demand, task durations, prices, resource availability) A user-friendly toolkit as plug-in to IBM Decision Optimization Center Cross-industry applications26 2015 IBM Corporation
5 steps Uncertainty Toolkit process1. Definedecision model 2. Characterizeuncertainty Create optimization model with IBM CPLEX StudioSome modeling skill required, or existing assetsEmbed in IBM Decision Optimization Center“Steve” the IT expert, &“Keith” the OR consultantOR consultant’s “wizard”: 7 screensDefines uncertainty, scenario generation, risk measures3. Generateuncertain model Built-in automated reformulation, based on steps 1 and 2No modeling knowledge required“Robustification” (make the original model robust to change)4. Generatedecisions Business user’s “wizard”Automated solution generationAutomated scenario comparison5. Analyzetrade-offs “Anne” the business userBuilt-in visual analyticsAnalyze KPI trade-offs across multiple plans & scenarios2727 2015 IBM Corporation
Uncertainty Toolkit: automated reformulationsRobust / Stochastic approachApplicablemodel typesResulting ingle-stage penalty approach(Mulvey et al., 1995)LPMILPLP (or QP)MILP (or MIQP)Scenarios (finite)No uncertain data in objectivefunctionTwo-stage penalty approach(Mulvey et al., 1995)LPMILPLP (or QP)MILP (or MIQP)Scenarios (finite)No uncertain data in objectivefunctionMultistage Stochastic(e.g. King & Wallace, 2012)LPMILPLPMILPLPMILPQCPMIQCPScenarios (finite)NoneRangeNo uncertain data in standaloneparameters orequality constraintsSafety margin approach with polyhedraluncertainty sets(Bertsimas & Sim, 2004)LPMILPLPMILPRangeNo uncertain data in standaloneparameters or equality constraintsExtreme Scenario approach(Lee, 2014)LPMILPLPMILPLPMILPLPMILPRangeNo uncertain data in variablecoefficientsScenariosUncertainty in standalone parametershandled as penalty term in objectiveSafety margin approach with ellipsoidaluncertainty sets(Ben-Tal & Nemirovski, 1999)Distributionally robust reformulation(Mevissen et al., 2013)Industry Solutions Joint Program28 2015 IBM Corporation
Uncertainty Toolkit: automated reformulationsRobust / Stochastic approachApplicablemodel typesResulting ingle-stage penalty approach(Mulvey et al., 1995)LPMILPLP (or QP)MILP (or MIQP)Scenarios (finite)No uncertain data in objectivefunctionTwo-stage penalty approach(Mulvey et al., 1995)LPMILPLP (or QP)MILP (or MIQP)Scenarios (finite)No uncertain data in objectivefunctionMultistage Stochastic(e.g. King & Wallace, 2012)LPMILPLPMILPLPMILPQCPMIQCPScenarios (finite)NoneQ: How do I know which of these methods to use?Safety margin approach with ellipsoidaluncertainty sets(Ben-Tal & Nemirovski, 1999)RangeNo uncertain data in standaloneparameters orequality constraintsA: The Uncertainty Toolkit will decide automatically based on your inputSafety margin approach with polyhedralLPinto the LPNo uncertain data in standaloneConsultant’sRangeWizarduncertainty sets(Bertsimas & Sim, 2004)MILPMILPExtreme Scenario approach(Lee, 2014)LPMILPLPMILPLPMILPLPMILPDistributionally robust reformulation(Mevissen et al., 2013)parameters or equality constraintsRangeNo uncertain data in variablecoefficientsScenariosUncertainty in standalone parametershandled as penalty term in objectiveIndustry Solutions Joint Program29 2015 IBM Corporation
Uncertainty Toolkit Decision Tree (automated)Select the uncertain data item(s)(1)Is the uncertainty represented
Business Analytics and Optimization Optimization From the book “Competing on Analytics” by Thomas H. Davenport , Jeanne G. Harris Standard Report Ad hoc reports Query/ Drill Down Alerts Statistical Analysis Forecasting/ Extrapolation Predictive Modeling ge Complexity Stochastic Optimization Descriptive Predictive Prescriptive
