MODEL:
! Newsvendor model with Endogenous Randomness, i.e., our decision
variables affect the distribution of the random variables.
Essential idea:
Stage 0: Choose parameters of the distribution.
Stage 1 beginning: Generate a standard random variable.
Stage 1 end: Transform standard r.v. to desired r.v. with
chosen parameters.
This specific case:
We can invest in advertising to increase mean demand.
We can invest in forecasting to decrease standard deviation in demand.
So our decisions in stage 0 affect the distribution of demand
seen in stage 1;
! Keywords: Advertising, Forecasting, Endogenous uncertainty, Uncertainty,
Stochastic optimization;
! Core model ---------------------------------------------------+;
DATA:
COST = 30; ! Purchase cost per unit;
SELLAT = 60; ! Selling price per unit;
COSTDUMP = 10; ! Cost/unit of disposing excess units;
COSTSHORT = 5; ! Shortage penalty per unit;
! Possible mean demands....;
MU0 = 60; MU1 = 61; MU2 = 62;
!as a function of amount spent on advertising;
CMU0 = 0; CMU1 = 25; CMU2 = 70;
! Possible standard deviations in forecast error...;
SD0 = 12; SD1 = 11; SD2 = 10;
!as a function of amount spent on forecasting, surveys, etc.;
CSD0 = 0; CSD1 = 14; CSD2 = 40;
SAMPSIZE = 10; ! Sample size;
ENDDATA
! Core model;
[R_OBJ] MAX = PROFIT;
PROFIT = SALES - COST * Q - DISPOSAL_COSTS - SHORTAGE_COSTS
- INVADVRT - INVFCAST; ! Minus investment in advertising & forecasting;
SURP - SHORT = Q - FDEMAND; ! Compute units over and under demand;
DISPOSAL_COSTS = COSTDUMP * SURP;
SHORTAGE_COSTS = COSTSHORT * SHORT;
SALES = SELLAT * ( FDEMAND - SHORT);
! Mean demand. The WMUi choose a point on the piecewise
linear curve;
DMUADJ = WMU0*MU0 + WMU1*MU1 + WMU2*MU2;
! Amount invested in advertising;
INVADVRT = WMU0*CMU0 + WMU1*CMU1 + WMU2*CMU2;
1 = WMU0 + WMU1 + WMU2;
! Effect of forecasting investment;
DSDADJ = WSD0*SD0 + WSD1*SD1 + WSD2*SD2;
INVFCAST = WSD0*CSD0 + WSD1*CSD1 + WSD2*CSD2;
1 = WSD0 + WSD1 + WSD2;
! Final demand, taking into account advertising and forecasting,
convert the standard random variable, DEMAND, to endogenous r.v.;
FDEMAND = DMUADJ + DSDADJ*DEMAND;
@FREE( DEMAND);
@FREE( FDEMAND);
! SP related declarations --------------------------------------+;
! 1) Staging information for decision variables;
! Initial decision is the order quantity at stage 0;
@SPSTGVAR( 0, Q); ! Quantity stocked;
@SPSTGVAR( 0, INVADVRT); ! Investment in advertising;
@SPSTGVAR( 0, INVFCAST); ! Investment in forecasting;
@SPSTGVAR( 0, DMUADJ); ! Adjusted mean demand;
@SPSTGVAR( 0, DSDADJ); ! Adjusted standard deviation in demand;
! Demand is a random variable in stage 1;
@SPSTGRNDV( 1, DEMAND);
! 2) Distributions;
! Generic demand is Normal 0, 1;
@SPDISTNORM( 0, 1, DEMAND);
! 3) Number scenarios or sample size;
@SPSAMPSIZE( 1, SAMPSIZE);
END
|