applications of lp/ilp models
DESCRIPTION
Applications of LP/ILP Models. Tran Van Hoai Faculty of Computer Science & Engineering HCMC University of Technology. Evolution of LP/ILP. Millions USD saved from applying LP/ILP to business/government Motivated by Simplex Digital computers Applied to. Building good LP/ILP models. - PowerPoint PPT PresentationTRANSCRIPT
Tran Van Hoai 1
Applications of LP/ILP Models
Tran Van HoaiFaculty of Computer Science & Engineering
HCMC University of Technology
2010-2011
Tran Van Hoai 2
Evolution of LP/ILP• Millions USD saved from applying LP/ILP to
business/government• Motivated by– Simplex– Digital computers
• Applied to
2010-2011
Aircraft fleet assignment
Health care Fire protection Diary production
Telecom network expansion
Bank portfolio selection
Defense/aerospace contracting
Military deployment
Air pollution control
Agriculture Land use planning
Tran Van Hoai 3
Building good LP/ILP models
• Familiarity– Limited resources– Overall (tradeoff) objective– Different perspectives
• Simplification– Models always simplify real-life, but which is
simplified is important• Clarity– Model must be clear
2010-2011
What constitutes the proper simplification is subject to individual judgment and experience(George Dantzig)
Tran Van Hoai 4
Summation variables/constraints
• Introduce new variables to be easier to understand/debug– Summation of variables/constraints
2010-2011
Production of 3 TV models• resource: 7000 pounds plastic
2 pounds/TV1, 3 pounds/TV2, 4 pounds/TV3• profit:
$23/TV1, $34/TV2, $45/TV3• management constraint: not any TV model exceed 40% total production
Tran Van Hoai 5
First modelMAX 23X1 + 34X2 + 45X3
S.T. 2X1 + 3X2 + 4X3 ≤ 7000
X1 ≤ .4(X1+X2+X3)
X2 ≤ .4(X1+X2+X3)
X3 ≤ .4(X1+X2+X3)
X1,X2,X3 ≥ 0
2010-2011
MAX 23X1 + 34X2 + 45X3
S.T. 2X1 + 3X2 + 4X3 ≤ 7000
.6X1 - .4X2 - .4X3 ≤ 0
-.4X1 + .6X2 - .4X3 ≤ 0
-.4X1 - .4X2 + .6X3 ≤ 0
X1,X2,X3 ≥ 0
No meaning as natural input(especially on spreadsheet)
Not management
constraint anymore
Tran Van Hoai 6
• Define summation variableX4 = total production of TVs
• Add summation constraintX1 + X2 + X3 - X4 = 0
MAX 23X1 + 34X2 + 45X3
S.T. 2X1 + 3X2 + 4X3 ≤ 7000X1 + X2 + X3 - X4 = 0X1 - .4X4 ≤ 0
X2 - .4X4 ≤ 0X3 - .4X4 ≤ 0
X1,X2,X3 ≥ 0
Revised model
2010-2011
Summation variable
Summation constraint
Clarity(although more variables/constraints)
Tran Van Hoai 7
Applications of LP/ILP
• More realistic example• Reduced version in different practical
applications– Portfolio model
2010-2011
Tran Van Hoai 8
Financial portfolio model
• Consider return projections of investment– Measure of risk,
volatility, liquidity, short/long term
• Highly nonlinear in nature, but we consider a linear case
2010-2011
Tran Van Hoai 9
Jones investment service(advise clients on investment)
2010-2011
Potential investment Expected return
John’s rating
Liquidity analysis
Risk factor
Savings account 4.0% A Immediate 0
Certificate of deposit 5.2% A 5-year 0
Atlantic Lighting 7.1% B+ Immediate 25
Arkansas REIT 10.0% B Immediate 30
Bedrock insurance annuity 8.2% A 1-year 20
Nocal mining bond 6.5% B+ 1-year 15
Minicomp systems 20.0% A Immediate 65
Antony hotels 12.5% C Immediate 40
Problem summary• Determine amount to be placed in each investment• Minimize total risk• Invest all $100,000• Meet the goals developed with client
annual return at least 7% at least 50% in A-rated investments at least 40% in immediately liquid investments no more $30,000 in savings and deposit
Tran Van Hoai 10
LP modelMIN 25X1 + 30X4 + 20X5 + 15X6 + 65X7 + 40X8
S.T. X1 + X2 + X3 + X4 + X5 + X6 + X7 + X8 = 100000
.04X1 + .052X2 + .71X3 + .1X4 + .082X5 + .065X6 + .2X7 + .125X8 ≥ 7500
X1 + X2 + X5 + X7 ≥ 50000
X1 + X3 + X4 + X7 + X8 ≥ 40000
X1 + X2 ≤ 30000
All Xi ≥ 0
2010-2011
Tran Van Hoai 11
Analysis (1)
• Binding constraints
• What exceed minimum requirements
• Investment
2010-2011
Expected annual return $7500Liquid investment $40000Savings & certificate deposit $30000
A-rated investment $50000 $77333
Savings $17333 Arkansas $22666Certificate deposit
$12666 Bedrock insurance
$47333
Tran Van Hoai 12
Analysis (2)• Reduced costs: in order to be included, risk
factor must be lowered
• Optimality range
• Shadow price: risk increased by
2010-2011
Atlantic 4.67 Minicomp 1.67Nocal 0.67 Antony 1.67
Bedrock 19.5 (20-0.5) -> 20.43 (20+0.43)Savings 0 -> 1.17
$1 extra to $100000 -7.33$1 extra liquid investments 4