1/48
Looks like no tags are added yet.
Feasible Region
location of all possible combinations of decision variables (x1, x2, etc) that satisfy the constraints
Optimization
Finding the point within the feasible region that maximizes or minimizes the objective function
Objective Function
A linear function comprised of two decision variables that is to be maximized or minimized according to the constraints.
Moves in a constant direction on the graph to intersect with the feasible region
Optimal value
The maximized or minimized value, it occurs at one of the extreme points (corner points) of the feasible region
How to find the Optimal Solution
1. Identify all corner points. 2. Compute the objective function at each point. 3. Pick the point with the optimal value
Infeasible Problem
no optimal solution exists —> No feasible region
How can we get multiple Optimal Solutions?
When the objective function has the same slope as one of the constraint boundaries.
Objective line slides along the boundary edge, producing multiple optimal points
All points on that boundary yield the same optimal value
Unbounded Problem
Occurs when the feasible region extends indefinitely in the direction of optimization
Indicates one or more constraints are missing or incorrectly specified
Sensitivity Analysis
Studies how changes in the parameters of an optimization model affect the optimal solution
Tests the robustness of the solution
Helps understand how much an input can change before the current solution is no longer optimal
Binding Constraint
A limit that the optimal solution satisfies precisely, leaving no slack.
Slack = 0 —> all available resources are fully used
Non-binding Constraint
A limit that has leftover capacity or unused resources, leaving slack
Slack > 0 —→ Changing RHS (right hand side) formula (within allowable range), without changing the objective value
Shadow price = 0 (sufficient condition)
Shadow Price
The change in the optimal objective value per 1-unit increase in RHS of a binding constraint
All Integer Optimization
all decision variables must be integers
Mixed Integer Optimization
some variables are integers, others are continuous
Binary Integer
variables take any values 0 or 1
LP Relaxation
integer restriction is removed to find approximate solution
Integer Programming
specialized algorithms that find optimal integer solutions
Linear Optimization
changes in inputs produce proportional changes in outputs
efficient and scalable
Nonlinear Optimization
Changes in input are not proportional
may have multiple local optima, which makes it harder to solve
local optimum
feasible solution where no nearby feasible point provides a better objective value
Local maximum
no neighboring feasible solution gives a LARGER objective value
Local minimum
no neighboring feasible solution gives a SMALLER objective value
global optimum
feasible solution where no other feasible solution within the feasible region gives a better objective value
Global maximum
no other solution gives a LARGER objective value
Global minimum
no other solution gives a SMALLER objective value
Concave function
bowl-shaped down, used in maximization problems
Convex function
bowl-shaped up, used in minimization problems
GRG Nonlinear Solver Method
searches locally for an optimum
suitable for smooth, differentiable nonlinear problems
Multistart Solver Option
Helps identify global optimum
runs solver from multiple starting points
Solver generates random starting values within the variable bounds
reports the objective value after testing all runs
Location model
determines where to place one or more facilities to achieve a desired objective
continuous location model
facility can be placed anywhere within a set space
discrete location model
facility must be chosen from a finite set of candidate sites
Median models
minimize the total weighted distance
Center models
minimize the maximum distance
Location Covering Models
Maximize the geographical amount of demand served within a target radius
Forecasting error
The difference between the actual observed value and the forecasted value
Product diffusion
the process of a new product spreading through the market over time
Bass Forecasting Model
describes how new products are adopted over time
Innovator
behavioral variable that is influenced by external factors
Imitator
behavioral variables that is influenced by social contagion or work-of-mouth
Portfolio Optimization Problem
determines how to allocate wealth across a set of assets, with the goal of choosing investment weights (how much to put into each investment)
Markowitz Portfolio Model
balances risk and return, with the objective of minimizing portfolio risk (variance)