alexei a. gaivoronski norwegian university of science and technology

33
Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 1 Stochastic optimization and modeling of network risk and uncertainty: the case of telecommunication services Alexei A. Gaivoronski Alexei A. Gaivoronski Norwegian University of Science Norwegian University of Science and Technology and Technology Joint work with Josip Zoric, Denis Becker, Joint work with Josip Zoric, Denis Becker, Adrian Werner, Paolo Pisciella Adrian Werner, Paolo Pisciella

Upload: armani

Post on 06-Jan-2016

20 views

Category:

Documents


0 download

DESCRIPTION

Stochastic optimization and modeling of network risk and uncertainty: the case of telecommunication services. Alexei A. Gaivoronski Norwegian University of Science and Technology Joint work with Josip Zoric, Denis Becker, Adrian Werner, Paolo Pisciella. Risk adapted performance networks. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.20071

Stochastic optimization and modeling of network risk and uncertainty:

the case of telecommunication services

Alexei A. GaivoronskiAlexei A. GaivoronskiNorwegian University of Science and Norwegian University of Science and

TechnologyTechnology

Joint work with Josip Zoric, Denis Becker, Adrian Werner, Joint work with Josip Zoric, Denis Becker, Adrian Werner, Paolo PisciellaPaolo Pisciella

Page 2: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 2

Risk adapted performance networksRisk adapted performance networks

Electric power generation and distributionElectric power generation and distribution Gas production, transportation, dirstributionGas production, transportation, dirstribution Telecommunications and internetTelecommunications and internet TransportationTransportation1.1. Hierarchical networks with nodes of different levels of Hierarchical networks with nodes of different levels of

complexity: from equipment to enterprisescomplexity: from equipment to enterprises

2.2. Nodes designed to meet local risk adjusted performance Nodes designed to meet local risk adjusted performance targets targets locallylocally

3.3. Network should satisfy risk/performance tradeoff Network should satisfy risk/performance tradeoff globallyglobally

4.4. Inherent uncertaintyInherent uncertainty

Page 3: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 3

Quantitative evaluation of business models for Quantitative evaluation of business models for collaborative service provisioncollaborative service provision

Work directionsWork directions

Getting qualitative understanding of business models, Getting qualitative understanding of business models, input from qualitative part, SPICE scenarios, surveysinput from qualitative part, SPICE scenarios, surveys

Development of quantitative modelsDevelopment of quantitative models Implementation in a prototype of decision support systemImplementation in a prototype of decision support system Testing on SPICE scenarios, casesTesting on SPICE scenarios, cases Deliverable on quantitative evaluationDeliverable on quantitative evaluation

Page 4: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 4

Status Task 1.4, quantitative analysis of business modelsStatus Task 1.4, quantitative analysis of business models

The Edition 1 of the set of models for investment business The Edition 1 of the set of models for investment business analysis of collaborative service provision has been analysis of collaborative service provision has been developed: top static viewdeveloped: top static view

Architecture of the prototype of decision support system Architecture of the prototype of decision support system for analysis of business models is selectedfor analysis of business models is selected

Parts of this prototype is under implementationParts of this prototype is under implementation

Page 5: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 5

Work in progressWork in progress

Edition 2 of the model set: service lifetime, different Edition 2 of the model set: service lifetime, different constellations of actorsconstellations of actors

Build up of the prototype of decision support system for Build up of the prototype of decision support system for business analysisbusiness analysis

Analysis of SPICE scenarios using the model setAnalysis of SPICE scenarios using the model set Analysis of possible business models using qualitative Analysis of possible business models using qualitative

input from other participantsinput from other participants Further dissemination effortFurther dissemination effort

Page 6: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 6

Quantitative business modelsQuantitative business models

What is it?What is it? Well understood in theory of corporate finance and in Well understood in theory of corporate finance and in

business practicebusiness practice BUT focus is on one single enerprise who selects industrial BUT focus is on one single enerprise who selects industrial

project or project portfolioproject or project portfolio Identify and measure and commeasure all cash flows Identify and measure and commeasure all cash flows

related to a given business activityrelated to a given business activity Give integrated assessment of cash flow/profit performance Give integrated assessment of cash flow/profit performance

based on different business principlesbased on different business principles Return on investment NPV Risk/performance tradeoff

Decision about business activityDecision about business activity Recent emphasis on risk controlRecent emphasis on risk control

Page 7: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 7

Challenges to this viewChallenges to this view

Networked industrial environmentNetworked industrial environmentDifferent independent agents are contributing to Different independent agents are contributing to

the common goal being in complex relations of the common goal being in complex relations of competition and collaborationcompetition and collaboration

How all this functions in such networked How all this functions in such networked environment?environment?

Corporate finance theory needs further Corporate finance theory needs further development for this casedevelopment for this case

Risk control issuesRisk control issuesGood example: evaluation of business models in Good example: evaluation of business models in

context of SPICEcontext of SPICE

Page 8: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 8

ObjectiveObjective

Starting from theory of corporate finance and Starting from theory of corporate finance and optimal decisions under uncertainty and risk optimal decisions under uncertainty and risk develop methods and tools for quantitative develop methods and tools for quantitative evaluation of business models in networked evaluation of business models in networked environmentenvironment

Utilize this methodology in SPICE context for Utilize this methodology in SPICE context for evaluation of collaborative service provision on evaluation of collaborative service provision on SPICE platformSPICE platform

Page 9: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 9

Risk/performance networksRisk/performance networks

State of the art: different attempts but no State of the art: different attempts but no universally accepted answersuniversally accepted answers

Growing importance in different fieldsGrowing importance in different fields Telecommunications Supply chain management Energy

Page 10: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 10

Example of structural description of service provision

Page 11: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 11

Different constellations of rolesDifferent constellations of roles

Page 12: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 12

Service architectureService architecture

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

Page 13: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 13

Services, roles and actorsServices, roles and actors

users services Components, enablers, roles

SPICE

actors

Page 14: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 14

Economic requirementsEconomic requirements

Platform should be attractive for all actorsPlatform should be attractive for all actorsActors should feel incentive to join service Actors should feel incentive to join service

provision, that is they should want to join provision, that is they should want to join cooperative effort because they will benefit from itcooperative effort because they will benefit from it

Services should provide to actors a competitive Services should provide to actors a competitive source of profitsource of profit

Risk/return considerations: risk that users will not Risk/return considerations: risk that users will not accept the service as expected, cannibalizing, etcaccept the service as expected, cannibalizing, etc

Page 15: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 15

Approach of modern financial theoryApproach of modern financial theory

Actors participate in service(s) provision assuming roles Actors participate in service(s) provision assuming roles and providing components for servicesand providing components for services

Quantify cash flow, profits and risksQuantify cash flow, profits and risks Each actor will select tradeoff between profit and risk Each actor will select tradeoff between profit and risk

exposure according to its preferencesexposure according to its preferences This will result in This will result in service portfolioservice portfolio for each actor for each actor Coordination tools should assure that the actors will select Coordination tools should assure that the actors will select

on their own accord participation in service provision in on their own accord participation in service provision in required proportionrequired proportion

Page 16: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 16

Risk/return tradeoffRisk/return tradeoffNobel prise winning conceptNobel prise winning concept

risk

return

feasible set

R

x

efficient frontier

x0x1

x2

Page 17: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 17

Quantitative modelQuantitative modelDescription of serviceDescription of service

Services consist of components which my be provided by Services consist of components which my be provided by different actorsdifferent actorsN components indexed by i and M services indexed by j

ij - share of component i in service j.

Description of service through components:

Service generate revenue Service generate revenue vvjj

Revenue sharing coefficients

Actor who contribures with component i recieves revenue

j 1j , . . , Nj

j 1j , . . , Nj

ijv i

Page 18: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 18

Description of actorsDescription of actors

Actors assume roles by providing service componentsActors assume roles by providing service components This incurs costs and brings revenueThis incurs costs and brings revenue

K actors indexed by k

cik – unit provision costs for actor k providing component i

Wik – provision capability of component i by actor k

xijk – the portion of provision capability for component i of actor k dedicated to participation in provision of service j.

Profit model for actor k

xijkWik - the volume of provision of component i dedicated by actor k to service j

Page 19: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 19

Profit model for actor Profit model for actor kk

xijkWik/λij- volume of service j in which the actor k participates

vjxijkWki/λij - the total revenue from this service

vjxijkWkiγij/λij - the part of the revenue which goes to actor k

Profit of actor k:

k j 1

M

v jx ijkWik ij ij

x ijkc ikW ik j 1

M

x ijkWikc ikv j ijcik ij

1

Page 20: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 20

Profit model for actor Profit model for actor kk

Basic case: an actor provides only one Basic case: an actor provides only one componentcomponent Profit

Return

Portfolio viewpoint: an actor chooses portfolio of services to which contribute

i Wic i j 1

M

x ijv j ijci ij

1

r i j 1

M

x ijv j ijci ij

1

x i x i1, . . . ,x iM

Page 21: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 21

Portfolio viewpointPortfolio viewpoint

Return coefficients associated with participation in each service

expected return coefficients

expected return

Risk that actual return will be different from Risk that actual return will be different from expected return or even become lossexpected return or even become loss

r ij v j ijci ij

1

ij ijEv jci ij

1

r i j 1

M

ijx ij j 1

M

x ij ijEv jci ij

1

Page 22: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 22

Efficient service portfoliosEfficient service portfolios

Problem to solve for computing eficient frontierProblem to solve for computing eficient frontier

minx StDev2 j 1

M

x ijv j ijci ij

1

j 1

M

x ij ijEv jci ij

1

j 1

M

x ij 1, x ij 0

Page 23: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 23

Next level: quantitative coordinationNext level: quantitative coordination

What is necessary is that the whole service What is necessary is that the whole service provision platform functions properlyprovision platform functions properly

And this means that different actors should And this means that different actors should independently make decisions to participate in independently make decisions to participate in different services which nevetherless will provide different services which nevetherless will provide coordinated result.coordinated result.

Revenue sharing coefficients should be chosen in Revenue sharing coefficients should be chosen in order to achieve thisorder to achieve this

Page 24: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 24

Coordinator (service provider) Coordinator (service provider) problemproblem

Paper is available on Edition 1 of the model set

Page 25: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 25

Architecture of the DSS prototypeArchitecture of the DSS prototype

Mathematicalmodel

Top level algorithmsScenario generation

Postprocessing

Problem solvers

Data and userinterface

Data User interaction

Results presentation

Excel MATLAB

XPRESS

SQG

CPLEX

results

data

User intervention

Service modelDetailed service

structure, resources

Service description

Page 26: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 26

Screenshot 1 of demo of DSS prototypeScreenshot 1 of demo of DSS prototype

Page 27: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 27

Screenshot 2 of demo of DSS prototypeScreenshot 2 of demo of DSS prototype

Page 28: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 28

Example: business person on the Example: business person on the movemove

Page 29: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 29

Risk/performance preferencesRisk/performance preferences

0.05

0.1

0.15

0.2

0.25

0.3

0 0.2 0.4 0.6 0.8

risk

pro

fit

Page 30: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 30

Market sharesMarket shares

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8

risk

pla

tfo

rm s

erv

ices

Page 31: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 31

Price competitionPrice competition

0

0.2

0.4

0.6

0.8

1

0.18 0.2 0.22 0.24 0.26 0.28 0.3 0.32 0.34

risk

pla

tfo

rm s

ervi

ces

-10%

-5%

+5

10

Page 32: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 32

SummarySummary

Modern theory of corporate finance and risk Modern theory of corporate finance and risk management together with optimization under management together with optimization under uncertainty provides a foundation for quantitative uncertainty provides a foundation for quantitative analysis of risk/performance networks in the analysis of risk/performance networks in the context of collaborative service provisioncontext of collaborative service provision

Page 33: Alexei A. Gaivoronski Norwegian University of Science and Technology

Alexei A. Gaivoronski IIASA, Workshop on Coping with Uncertainty, 10-12.12.2007 33

ConclusionsConclusions

Traditional risk management paradigm should be augmented and Traditional risk management paradigm should be augmented and developed further: noncommeasurable risksdeveloped further: noncommeasurable risks

Modern theory of corporate finance and risk management provides a Modern theory of corporate finance and risk management provides a foundation for quantitative analysis of risk/performance networks but foundation for quantitative analysis of risk/performance networks but much more work is neededmuch more work is needed

Many possibilities for stochastic programming approachesMany possibilities for stochastic programming approaches Three components: Modern computing technology, off-shelf Three components: Modern computing technology, off-shelf

optimization software, custom algorithm designoptimization software, custom algorithm design It is possible to solve highly nonlinear and nonconvex problems in It is possible to solve highly nonlinear and nonconvex problems in

industrial quantitiesindustrial quantities