Unit III

LINEAR PROGRAMMING:
SPECIAL-PURPOSE ALGORITHMS

INTRODUCTION

More efficient computational procedures than the simplex method can be used in
solving some types of linear programming problems. One of the most useful of these
special-purpose algorithms is the transportation method examined in this unit. The
assignment problem, a special case of the more general transportation problem, will also
be discussed.

The transportation problem is concerned with selecting different routes between

manufacturing plants and distribution warehouses or between regional distribution
warehouses and local distribution outlets. The assignment problem, on the other hand,
involves assigning employees to tasks, salespersons to territories, contracts to bidders, or
jobs to plants. The transportation method and assignment method can be applied by
management in searching for a distribution route or an assignment which will optimize
some objective; this can be the minimization of total transportation cost, the
maximization of profit, or the minimization of total time involved.

SESSION OBJECTIVES

Introduce some of the special-purpose algorithms used to solve particular kinds of

linear programs.
Provide specific instruction in the use of the transportation method, including both
balanced and unbalanced conditions of supply and demand.
Introduce the assignment method for a special class of assignment problems and
demonstrate its use.
Show to handle degeneracy in both the transportation and assignment methods.

SUGGESTED TIMEFRAME: 6 hours

THE TRANSPORTATION METHOD (DEMAND EQUALS SUPPLY)

The transportation method was first formulated as a special procedure for finding the
minimum cost program for distributing homogeneous units of a product from several points of
supply (source) to a number of points of demand (destinations). Suppose a manufacturer has 5
plants (source) and 20 warehouses (destinations), all located at different geographical points. For a
specified time, each source has a given capacity and each destination has a given requirement, and
the costs of shipping the product from each source to each destination are known. The objective is
to schedule shipments from sources to destinations in such a way as to minimize the total
transportation cost.

The earliest formulation of this basic transportation problem was stated by F.L. Hitchcock
in 1941 and later expanded by T. C. Koopmans. The linear programming formulation was first
given by G. B. Dantzig. In 1953, W. W. Cooper and A. Charnes developed the stepping-stone
method, a special-purpose algorithm for solving the transportation problem. Subsequent
improvements led to the computationally easier modified distribution (MODI) method in 1995.

Let us consider the case of the Bali Gravel Supply, which has received a contract to supply
gravel for three new house projects located in the subdivisions of Ville A, Ville B, and Ville C.
Construction engineers have estimated the amounts of gravel which will be needed at three house
construction projects:

Weekly
requirement
Project Location (truckloads)
A Ville A 72
B Ville B 102
C Ville C 41
Total 215

The Bali Gravel Supply has three gravel plants located in the towns of Kervin, Wervin,
and Bervin. The gravel required for the construction projects can be supplied by these
three plants. Balis chief dispatcher has calculated the amounts of gravel which can be
supplied by each plant:

Amount
available/week
Plant Location truckloads
W Kervin 56
X Wervin 82
Y Bervin 77
Total available 215
 The total amount available is exactly equal to the total amount required. When
total supply is equal to total demand, a balanced condition is said to exist. Although the
balanced case is very unlikely in actual practice, it will enable us to focus on the basic
ideas underlying the transportation method.

The company has computed the delivery costs from each plant to each project
site. As in the linear programming problems discussed in the previous chapter, we assume
that the variables in the problem must be linearly related. In this case, delivery costs per
truckload between each plant and project site vary directly with the quantity distributed.
These costs are shown in Table 3-1.

In Figure 3-1, we have illustrated this problem graphically. The circles represent
the projects, and the rectangles represent the gravel pits. Delivery costs per truckload are
shown on the arrows which connect each plant with each house construction project.

Given the amounts required at each project and the amounts available at each
plant, the companys problem is to schedule shipments from each plant to each project in
such a manner as to minimize the total transportation cost within the constraints imposed
by plant capacities and project requirements. At this point we have all the information
necessary to solve Balis problem.

Table 12-1
DELIVERY COSTS

Cost per truckload

From To project A To project B To project C
Plant W $ 4 $ 8 $ 8
Plant X 16 24 16
Plant Y 8 16 24
Figure 3-1
Gravel plants, road
construction projects,
and transportation
costs for Bali Gravel
Supply

STEP 1: SET UP THE TRANSPORTATION TABLEAU

The transportation tableau serves the same basic purpose as the simplex tableau; it provides a
framework for presenting all the relevant data in a concise manner and facilities the search for
progressively better solutions. In Figure 3-2 the standard format for the transportation tableau has
been divided into five lettered sections-A, B, C, D, and E- each of which will be explained in detail.

Section A In this part we list the source of supply, plants W, X, and Y. Each row represents a plant in
the tableau.

Section B The capacity for each plant is shown in section B. Thus we can think of the rows of the
tableau as representing the capacity constraints. These are also referred to as the rim requirements for
the rows. For example, the rim requirement in the first row means that plant W can supply no more
than 56 truckloads per week.

Section C The destination points are listed in this section, house construction projects A, B, and C.
Each project represents a column in the tableau.

Section D The requirements for each project are placed in this part. The columns, then, represent the
demand constraints, or rim requirements, for the projects. For example, the rim requirement for
column 1 signifies that project A requires exactly 72 truckloads per week. The total number of rim
requirements in our problem is six -- three for the rows and three for the columns.
Figure 3-2
Transportation
tableau for Bali
Gravel Supply
problem.

Section E In this section are nine squares, or cells, representing the alternative source-to-destination
assignments that could be made. For example, the 56 truckloads per week available at plant W may be
used, in whole or in part, to fulfill the requirements of any of the three projects. Any combination of
shipments from plant W would be acceptable as long as the total equaled exactly 56 truckloads. Similarly,
the 72 truckloads required at project A may be met by any combination of shipments from the various
plants as long as the total equals 72.

To complete the tableau, it will be helpful to add to each square of section E an identification symbol
and a delivery cost figure. This has been done in Figure 3-3.

Let us examine square WA located in the upper left-hand corner, repeated in Figure 3-4.

1. The WA in the upper left-hand corner of the square is the identification symbol. This square represents
the combination plant W to project A and therefore is identified as square WA.
2. The 4 in the upper right-hand corner is the transportation cost per truckload between plant W and
project A. The costs in each square were obtained from Table 3-1.
3. X1 represents the number of truckloads shipped from Bali plant W to project A. In other words, all Xs
in figure 3-3 denote the number of shipments between each plant and each project. The value of each X
will be a positive whole number or zero. If in a particular solution the X value is missing for a square,
this means that no quantity is shipped between the plant and project in question.
Figure 3-3
Transportation
tableau with squares
identified and
delivery costs added.

STEP 2: DEVELOP AN INITIAL SOLUTION

After arranging the data in tableau form, the next step is to find a solution to the problem in
order to provide a starting point leading to the procedure for developing improved solutions. Thus the
initial solution in the transportation method serves the same purpose as the initial solution in the
simplex method.

A systematic and logical procedure known as the northwest corner rule has been developed
for setting up the initial solution. Although it does not have to be used, this rule offers the advantage
of being systematic rather than trial and error. The northwest corner rule may be stated as follows:

1. Starting at the upper left-hand corner (the northwest corner) of the tableau, the supply available at

each row must be exhausted before moving down to the next

row, and the rim requirement of any column must be exhausted before moving to the right to the next
column.

2. Check to see that all rim requirements have been satisfied.

The results of this procedure are shown in Figure 3-5.

 An explanation of each assignment made in the initial solution shown in Figure 3-5 is given
below:

Figure 3-5. The initial solution to the Bali Gravel Supply problem.

Square WA Beginning in the upper left-hand corner, we compared the quantity available at plant W (56)
with the quantity required at project A (72). Exhausting the supply at plant W, 56 truckloads are shipped to
project A. This leaves project A short 16 truckloads. Move down to the second row in the same column to
square XA.

Square XA Plant X has 82 truckloads available. Since project A is 16 short, Bali plant X will ship 16 of its
82 available truckloads to project A. The requirements for project A have now been met. Since plant X has
66 truckloads remaining, we move to the right to the next column to square XB.

Square XB Project B needs 102 truckloads. The remaining 66 truckloads from plant X will then be shipped
to project B, leaving project B short 36 truckloads. As the amount available at plant X has been exhausted,
we move vertically down to the next row to square YB.

Square YB Plant Y has 77 truckloads available. Project B needs 36 more truckloads to fully satisfy its
requirements. Hence plant Y will ship 36 of its 77 available truckloads to project B. We now move to the
right to the next column to square YC.

Table 3-2 Initial Solution Illustrated

From plant To project Quantity, truckloads/week

W A 56
X A 16
X B 66
Y B 36
Y C 41
Total 215
Square YC Plant Y has 41 truckloads remaining, which are shipped to project C,
requiring 41 truckloads. The schedule of shipments is now complete.

The initial solution, then, includes the five source-destination combinations

shown in Table 3-2. The squares in Figure 3-5 where no circled values appear are
referred to as unused squares; that is, no quantity is shipped between the two points
represented by an unused square. These squares therefore are not in the initial
solution.
We must now determine the cost of this first solution for the Bali Gravel Supply.
To do this we multiply the quantities shipped between each source- destination
combination in the solution by the respective unit cost. The results are shown in
Table 3-3.
Before proceeding to the next step, several points should be made. The initial
solution is a feasible one as all rim requirements have been met; that is, the sum of
each row or column is equal to its rim requirement. Initial solutions obtained by
using the northwest corner rule can always be recognized by their stairstep
appearance, as shown by the path described in Figure 3-5.
Finally, for any solution, the number of used squares must be equal to the total
number of rim requirements minus 1. In our first solution, there are 5 used squares,

Table 3-3 Total Cost of Initial Solution

Source-destination Quantity x Unit = Total

Combination shipped cost cost

WA 56 $ 4 $ 224
XA 16 16 256
XB 66 24 1584
YB 36 16 576
YC 41 24 984
Total transportation cost $ 3624

or source-destination combinations. The total rim requirements are 6, 1 for each

plant (row) and project (column). Thus

Used squares = total rim requirements 1

5=61

When any solution does not conform to the above rule, degeneracy exists. (Note:
Initial solution is not degenerate.) The important point here is that each solution
should be tested for degeneracy; that is, the number of used squares must be equal
to the total rim requirements minus 1.
STEP 3: TEST THE SOLUTION FOR IMPROVEMENT

The evaluation procedure involves the examination of each unused square in the tableau to see
whether it is more desirable to move a shipment into one of them. Its purpose is to determine whether a
better schedule of shipments from plants to projects can be developed.

Two alternative procedures for evaluating the unused squares are the stepping-stone and MODI
methods. The stepping-stone method is the basis for the MODI method and provides a good
introduction to it.

The used squares, those containing circled values, are said to be in solution and will be referred
to as stone squares. In applying the stepping stone method, we ask this question: what would happen if
one truckload of gravel were tentatively shipped or assigned to an unused square? If this tentative
assignment results in a favorable effect (reduces cost), the unused square evaluated then becomes a
possible candidate for entering the next solution.

To apply this to our present problem, we have to take note that the square WB is unused in
Figure3-5. Suppose that we assign 1 truckload to square WB, that is, ship 1 truckload from plant W to
project B. In order to make this assignment and still satisfy the capacity restriction (rim requirement) for
plant W, we must subtract from square WA 1 truckload so that the total shipments from plant W do not
exceed 56. However, if we subtract 1 truckload from square WA, we must then add 1 truckload to
square XA in order to meet the rim requirement for project A.

Adding 1 truckload to square XA means that we must subtract 1 truckload from square XB in
order to satisfy the rim requirement for that row (plant X). Finally, the truckload subtracted from square
XB still enables project B requirements to total 102; one truckload has been tentatively added to square
WB at the start of the evaluation. The evaluation is now back where it started. Please see Figure 3-6.

Note that the net change for any row or column is zero; wherever 1 truckload was added to a
square, another square was decreased by the same amount.

Figure 3-6. Adjustment required in evaluating square WB.

 Looking at the path described in evaluating square WB (Figure 3-6), we see that shipping 1
truckload from plant X to project B results in an increase of $8 in distribution costs. The $8 cost per
truckload between the two points is given in the upper right-hand corner of square WB. A similar
increase from plant X to project A results in additional cost of $16 (square XA). Likewise, the decrease
of 1 truckload between plant W and project A reduces total costs by $4 (square WA). In addition, the
decrease of 1 truckload between plant X and project B results in reduction of $24 (square XB). The net
change in costs, referred to as the improvement index, is computed as follows:

Addition to cost: From plant W to project B $8

From plant X to project A 16 $24

Reduction in cost From plant W to project A $4

From plant X to project B 24 28
-$4

The same answer can be obtained by following the path used directly and resorting to a sort of
shorthand, as follows:

Improvement index for square WB = WB WA + XA - XB

Now substituting the cost per truckload for each source-destination combination in the above equation,
we have

Improvement index for square WB = $8 - $4 + $16 - $24 = -$4

The -$4 means that for every truckload shipped from plant W to project B, total transportation
costs would be reduced by $4. Because this is true, it would be advantageous to use this route if this
were the only choice available. However, the evaluation of other unused squares in our tableau might
bring about an even greater reduction. The task remaining then is to evaluate all remaining unused
squares.

STEPS IN EVALUATING ANY UNUSED SQUARE

1. Choose the unused square to be evaluated.

2. Beginning with the selected unused square, trace a closed path (moving horizontally and vertically
only) from this unused square via stone squares back to the original unused square. Only one closed
path exists for each unused square in a given solution. Although the path may skip over stone or
unused squares, corners of the closed path may occur only at the stone squares and the unused square
being evaluated. Only the most direct route is used.

3. Assign plus (+) and minus (-) signs alternately at each corner square of the closed path, beginning
with a plus sign at the unused square. Assign these signs by starting in either a clockwise or
counterclockwise direction. The positive and negative signs represent the addition or subtraction of 1
unit (truckload in this case) to a square.
4. Determine the net change in costs as a result of the changes made in tracing the path. Summing the
unit cost in each square with a plus sign will give the addition to cost. The decrease in cost is
obtained by summing the unit cost in each square with a negative sign. Comparing the additions to
cost with the decreases will give the improvement index.

5. Repeat the above steps until an improvement index has been determined for each unused square.

If all the indices are greater than or equal to zero, an optimal solution has been found.
Conversely, if any of the indices is negative, an improved solution is possible.

Let us now evaluate the unused square WC. The traced path used in evaluating this square is
shown in Figure 3-7. The improvement index for unused square WC is computed as follows:

Improvement index for WC = WC WA + XA XB + YB YC

= $8 - $4 + $16 - $24 + $16 - $24
= -$12

The closed path and improvement indices for the remaining two unused squares are

Path for square XC; (+)XC (-)XB (+)YB (-)YC

Improvement index for XC = XC XB + YB YC
= $16 - $24 + $16 - $24
= -$16

Path for square YA; (+)YA (-)XA (+)XB (-)YB

Improvement index for YA = YA XA + XB YB
= $8 - $16 + $ 24 - $16
= $0

Our evaluation resulted in three unused squares with negative improvement indices, and so we
know that a better solution is possible. The next step, then, is to develop the new solution.

STEP 4: DEVELOP THE IMPROVED SOLTUION

Each negative improvement index represents the amount by which Balis total transportation
costs could be reduced if 1 truckload were shipped by that source-destination combination. For
example, the improvement index for square XC means that for every truckload shipped from plant X to
 Figure 3-7. Path used in evaluating square WC.

Figure 3-8. All unused squares evaluated.

project C, total transportation costs will be reduced by $16. The question now is: given three alternative
routes with negative improvement indices (squares WB, WC, and XC), which one shall we choose in
developing the improved solution? We shall select that route (unused square) with the largest negative
improvement index.. In our problem this is square XC, with a negative index of $16.

Using this route will reduce costs. Bali must now decide how many truckloads to ship via this
route, that is, from plant X to project C. To do this, we must reconstruct the closed path traced in
evaluating unused square XC, as shown in Figure 3-9, using only the relevant part of the tableau.

Now the maximum quantity which Bass can ship from plant X to project C is found by
determining the smallest stone in a negative position on the closed path. The closed path for square XC
has negative corners at squares XB and YC, and the smaller of these two stones is 41 truckloads per
week. To obtain our new solution, we add 41 truckloads to all squares on the closed path with plus
 Figure 3-9. Closed path traced in evaluating unused square XC.

Figure 3-10. Forty-one loads added and subtracted.

signs, and we subtract this quantity from all squares on the path assigned minus signs, as shown in
figure 3-10.

Why did we choose the smallest stone in a negative position on the closed path as the maximum
number of truckloads that could be shipped from plant X to project C? Suppose we add 66 truckloads to
square XC instead of 41. In order to satisfy our rim requirements, we would have to do the following;
add 66 to XC, subtract 66 from XB, add 66 to YB, and subtract 66 from YC. The results are

XC: 0 + 66 = 66
XB: 66 66 = 0
YB: 36 + 66 = 102
YC: 41 66 = -25

The computation for square YC shows that Bali Gravel Supply plant Y would ship -25
truckloads to project C. This negative shipment is both meaningless in an actual problem and violation
of the requirement prohibiting the assignment of negative stone squares. All values in the solution must
be greater than or equal to zero; hence the maximum quantity which may be brought into a solution is
found by determining the smallest stone in a negative position on the closed path of the square with the
largest negative improvement index. This quantity is added to all squares on the closed path with plus
signs and subtracted from all squares on the path with minus signs. The new improved solution is
shown in Figure 3-11.

Note that square YC, which was a stone square in the initial solution, is now an unused square.
Square XC has entered the improved solution in place of square YC. As can be seen in Table 3-4, the
total transportation cost for the new shipping assignments of our second solution is an improvement
upon the cost of the first solution, $2968 versus $3624.
 We now go back to step 3 to determine whether further improvement is possible. Using the
stepping-stone method described in that section, we calculate an improvement index for each unused
square in the second solution. The closed path and improvement index for each unused square in Figure
3-11 are shown in Table 3-5.

Figure 3-11. The second solution.

In Table 3-5, there is one negative index for unused square WB, indicating that further
improvement is possible. For each truckload assigned to square WB, total costs will be reduced $4. To
determine the number of truckloads to be shipped, we select the smallest stone in a negative position on
the closed path traced in evaluating square WB (step 4).

+ WB WA + XA XB + - 56 + 16 - 25

As seen above, the smallest stone in a negative position is 25. This quantity is added to all
squares on the path with plus signs and subtracted from all squares on the path with minus signs, as
shown in Figure 3-13.

Table 3-4 Total Cost of Second Solution

Shipping Quantity x Unit = Total

assignments shipped cost cost

WA 56 $4 $ 224
XA 16 16 256
XB 25 24 600
XC 41 16 656
YB 77 16 1232
Total transportation cost $2968
Table 3-5
Closed Paths and Improvement Indices for Unused Squares of Figure 3-11

Unused Closed Computation of

square path improvement index

WB +WB WA + XA XB +8 4 + 16 24 = -4
WC +WC WA + XA XC +8 4 + 16 16 = +4
YA +YA XA + XB YB +8 16 + 24 16 = 0
YC + YC XC + XB YB +24 16 + 24 16 = +16

The third improved solution is given in Figure 3-14, and the total cost of the third solution is
shown in Table 3-6.

We go back to step 3 again to determine if further improvement is possible. The closed paths and
improvement indices for unused squares in Figure 3-14 are given in Table 3-7. The negative
improvement index for square YA indicates that the best solution has not yet been obtained. Now,
following the same procedure discussed in step 4, we find that the maximum quantity to assign to
square YA is 31 truckloads;

Figure 3-12. Tracing the path for square WC.

Figure 3-13. Twenty-five loads added and subtracted.

 Figure 3.14. The third solution.

the new solution is given in Figure 3-15. This is indeed the optimum solution. The improvement
indices for Figure 3-15 are all greater than or equal to zero (see Table 3-8).

Table 3-6 Total Cost of Third Solution

Shipping Quantity x Unit = Total

assignments shipped cost cost

WA 31 $4 $ 124
WB 25 8 200
XA 41 16 656
XC 41 16 656
YB 77 16 1232
Total transportation cost $2868

Table 3-7 Closed Paths and Improvement Indices for Figure 3-4

Unused Closed Computation of

square path improvement index

WC +WC WA + XA XC +8 4 + 16 16 = +4
XB +XB WB + WA XA +24 8 + 4 16 = +4
YA +YA WA + WB YB +8 4 + 8 16 = -4
YC +YC YB + WB WA + XA XC +24 16 + 8 4 + 16 16 = +12
 Figure 3-15. The optimum solution to the Bali Gravel Supply problem.

The total cost for the optimum solution to the Bali Gravel Supply problem is shown in Table 3-9.
Bali might question the use of a somewhat tedious method for solving such a simple problem. Why not
use a trial-and-error method? Why not begin by simply choosing the lowest-cost route and using it

Table 3-8 Closed Paths and Improvement Indices for Figure 3-15

Unused Closed Computation of

assignment path improvement index

WA +WA YA + YB WB +4 8 + 16 8 = +4
WC +WC XC +XA YA + YB WB +8 16 + 16 8 + 16 8 = +8
XB +XB YB + YA XA +24 16 + 8 16 = 0
YC +YC XC + XA YA +24 16 + 16 8 = +16

Table 3-9 Total Cost Of Optimum Solution for Bali Gravel Supply Problem

Shipping Quantity x Unit = Total

assignments shipped cost cost

WA 56 $ 8 $ 448
WB 41 16 656
XA 41 16 656
XC 31 8 248
YB 46 16 736
Total transportation cost $2744

to the fullest extent? Then we might select the next highest rate and use it to the fullest extent, and so
on, until we have satisfied all the requirements for the project. The assumption underlying this thought
is that making the best possible choice in each part of the scheduling program automatically result in the
best overall program. An examination of the final solution to our problem indicates that the assumption
is not a valid one. Figure 3-15 shows that two of the least-cost routes, squares WA and WC are not
included in the optimal solution. Here we recognize the importance of a characteristic of operations
research: a problem must be studied in terms of the total system, not merely the separate parts. The
optimal source-destination combinations are often far from apparent, even in small problems like that of
the Bali Gravel Supply.

ALTERNATIVE OPTIMAL SOLUTIONS

In the optimal solution to the Bali Gravel Supply problem, the improvement index of square XB
is zero. What does this signify? A zero improvement index for an unused square means that if these
routes were brought into the solution, the shipping assignments would change, yet the total
transportation cost would be the same. Thus, if we were to assign one truckload to unused square XB,
the total cost figure would neither increase nor decrease. We can conclude, then, that in addition to our
present optimal shipping schedule, another equally profitable schedule exists. To determine what this
alternative optimal solution is, we follow the same procedure used for bringing any route into the
solution (step 4). In this case the maximum number of truckloads that can be assigned to square XB is
41. The alternative solution and its total cost are given in Figure 3-16. The improvement indices are also
included.

The total cost is exactly the same as the total cost of the original solution. Also, the improvement
indices are positive except for square XA, which equals to zero. This is to be expected, as square XA
was the one replaced in the original optimal solution. From a practical viewpoint, the existence of
alternative optimal solutions gives valuable flexibility to the management of the Bali Gravel Supply.
THE MODI METHOD FOR COMPUTING IMPROVEMENT INDICES

The modified distribution method, referred to as the MODI method, is very similar to the
stepping-stone method except that it provides a more efficient means for computing the improvement
indices for unused squares. The major difference between these two methods concerns that step in the
problem solution at which the closed paths are traced. In order to calculate the improvement indices for
a particular solution, it was necessary in the stepping-stone method to trace a closed path for each

Figure 3-16. Alternative solution and total cost.

unused square. The unused square with the most improvement potential (the largest negative value) was
then selected to enter the next solution.

In the MODI method, however, the improvement indices can be calculated without drawing the
closed paths. The MODI method, in fact, requires tracing only one closed path. This path is drawn after
the unused square with the highest improvement index has been identified. As in the stepping-stone
method, the purpose of this path is to determine the maximum quantity that can be assigned to the
unused square entering the next solution.

Using the Bali problem, we shall illustrate the procedures used in applying the MODI method.
Beginning with the same initial solution obtained by using the northwest corner rule, the first step is to
compute a value for each row and each column in the transportation tableau. These values depend on
the particular solution and are used to compute the improvement indices for the unused squares.
Assigning a number to each row and column requires a slight modification in the transportation tableau.
This modification together with the initial solution is shown in Figure 3-17.

In this figure, we let R and K represent the row and column values. We have attached a
subscript to denote the specific row and column value. In our case, we have R 1, R2, and R3 to represent
the rows and K1, K2 and K3 to represent the columns. In general, then, we can say
Ri = value assigned to row i and Kj = value assigned to column j

The transportation cost, as in previous tableaus, is shown in the upper right-hand corner, or
subsquare, of each large square. For identification purposes, we can let C ii = cost in square ij (the square
at the intersection of row i and column j).
 Figure 3-17. Transportation tableau using the MODI method.

For example, C12 represents the cost in the square located at the intersection of row 1 and column 2.
now to compute the values for each row and column, we use formula

Cost at stone square ij Cij = Ri + Kj

This formula is applied only to the stone squares in a particular solution. Because there are five stone
squares in our problem, we must have five equations. For the stone square located at the intersection of
row 1 and column 1, we write
R1 + K1 = C11
And as there is a stone at the intersection of row 2 and column 1,
R2 + K1 = C21
Similarly,
R2 + K2 = C22
R3 + K2 = C32
R3 + K3 = C33

Because we are given the cost figure for each square in the table, we can substitute the appropriate
value for each Cij in our equations. The results are
R1 + K 1 = 4
R 2 + K1 = 16
R2 + K2 = 24
R3 + K2 = 16
R3 + K3 = 24

Notice that we have six unknowns and only five questions. Thus this system of equations has several
solutions. In order to find a particular solution (a value for each R and K), we shall let R 1 = 0. We could
have chosen any value for any row or column, but the usual procedure is to let row 1 (R 1) equal to zero.
This is legitimate since the entire process is a comparative one. In other words, the significance of the
row column values is not absolute numerical value. We are interested only in comparing the figures, not
in the figures themselves.

To solve the five equations, then, we proceed as follows. If R1 = 0, then

 K1 + K1 = 4
0 + K1 = 4
K1 = 4

Since K1 = 4, then R2 + K1 = 16
R2 + 4 = 16
R2 = 12

Since R2 = 12, then R2 + K2 = 24

12 + K2 = 24
K2 = 12

Since K2 = 12, then R3 + K2 = 16

R3 + 12 = 16
R3 = 4

Since R3 = 4, then R3 + K3 = 24
4 + K3 = 24
K3 = 20

The R and K values need not always be positive; indeed, they may be positive, negative, or zero.
After some practice, computing the R and K values can usually be done mentally instead of writing out
each equation as above. The transportation tableau with the R and K values included is shown in Figure
3-18.

With the row and column values computed, the next step in the MODI method is to evaluate
each unused square in the present solution, that is, to compute the improvement indices. Computing the
improvement index for any unused square is accomplished in the following manner; from the cost of an
unused square subtract the corresponding row value and column value. Starting this rule as a general
formula, we have

Improvement index = Cij Ri - Kj

If the result is negative, further improvement is possible. When all indices are equal to or greater than
zero, the optimal solution has been obtained.

Each unused square in the initial solution (Figure 3-18) can now be evaluated. For example, the route
from plant W to project B at the intersection of row 1 and column 2 is one of the unused routes (unused
squares) in our initial solution. Using our formula, we have

Unused square 12: C12 R1 K2 = improvement index

8 0 12 = -4
 Figure 3-18. Initial solution with R and K values for the Bali Gravel Supply problem

Similarly, for the other unused squares we have;

Unused Improvement
Square Cij Ri Kj index

13 C13 R1 K3
8 0 20 -12
23 C23 R2 K3
16 12 20 -16
31 C31 R3 K1
844 0

A comparison of these improvement indices with those obtained using the stepping-stone
method (see figure 3-8) shows them to be identical. From this point on, then, the procedure for
developing a new, improved solution is identical to the one discussed in the previous sections.

Procedure for developing a new improved solution

1. Trace a closed path for the cell having the largest negative improvement index.
2. Place plus and minus signs at alternate corners of the path, beginning with a plus sign at the
unused square.
3. The smallest stone in a negative position on the closed path indicates the quantity that can be
assigned to the unused squares being entered into the solution. This quantity is added to all
squares on the closed path with plus signs and subtracted from those squares with minus signs.
4. Finally, the improvement indices for the new solution are calculated.

With this procedure, the second solution to the problem of the Bali Gravel Supply is
obtained (figure 3-9). Notice that it is identical to our second solution found by using the
stepping-stone procedure (figure 3-11). As we shall see in a moment, this holds true for all
solutions.
 To evaluate the unused squares of the second solution using the MODI method, we must
calculate the R and K values. This must be done with every new solution. Again we begin by
letting R1 equal zero. Using the general formula Ri + Kj = Cij (the cost at stone square ij), the R
and K values are computed as follows:

Stone square 11: R 1 + K1 = 4

0 + K1 = 4
K1 = 4
Stone square 21: R2 + K1 = 16
12 + K2 = 24
K2 = 12
Stone square 22: R2 + K2 = 24
12 + K2 = 24
K2 = 12
Stone square 23: R2 + K3 = 16
12 + K3 = 16
K3 = 4
Stone square 32: R3 + K2 = 16
R3 + 12 = 16
R3 = 4

These R and K values were included in Figure 3-9. in comparing our new R and K values
with those obtained in the initial solution (Figure 3-18), we find that all R and K values are the
same except for K3, which is equal to 4 in the second tableau. Changing to solution changes
some if not all of all of the R and K values. Hence, with every new solution, new values for R
and K must be established in order to determine whether further improvement is possible, that is,
to calculate the improvement indices.
YOUR ACTIVITY

1. You have begun a business of your own and have decided to produce one or more of
products A, B, C, and D. You have approached four banks W, X, Y, and Z with your ideas
on these projects in order to obtain the necessary financing. The following table reflects the
level of financing required for each project, the interest rate each of the banks is willing to
charge on loans for each of the projects, and the total line of credit each of the banks is
willing to lend you.

Project (interest rate) Maximum

Bank A B C D credit
W 165 185 19% 17% P200,000
X 15 17 20 16 100,000
Y 17 16 18 18 200,000
Z 18 19 19 18 300,000
Amount
required P400,000 300,000 200,000 200,000

As each project should be as attractive profitwise as any other, you have decided to undertake all
or part of any number of projects you can at the lowest total payment of interest. Which projects
should you adopt and from which banks should you finance them.

2. Ace Manufacturing Company has orders for three similar products:

Product Orders (units)

A 3,500
B 1,500
C 2,500

Three machines are available for the manufacturing operations. All three machines can produce
all the products at the same production rate. However, due to varying defect percentages of each
product on each machine, the unit costs of the products vary depending upon the machine used.
Machine capacities for the next week, and the unit costs, are as follows;

Machine Capacity (units)

I 2,500
II 2,500
III 2,500

Machine Product
A B C
I 10 12 9
II 13 14 12
III 11 10 12
 a. Use the transportation model to develop the minimum cost production schedule for the
products and machines.
b. Does an optimal solution exist? Please verify using the Stepping Stone Method and the
Modified Distribution (MODI) Method.
c. If the given unit costs are profit per unit, what will be the optimal solution of the
problem?

REFERENCES

Andersen, D.R., D.J. Sweeney, & T.A. Williams, An Introduction to Management Science:
Quantitative Approaches to Decision Making

Andersen, D. R., Dennis J. Sweeny, and Thomas A. Williams. 2001. Quantitative Methods for
Business. Eighth edition. Cincinnati, Ohio: South-Western College Publishing.

Anderson, Michael Q. 1982. Quantitative Management Decision Making, Belmont, California:

Brooks/ Cole Publishing Co.,

Dunn, Robert A. and K. D. Ramsing. 1981. Management Science: A Practical Approach to

Decision Making. New York, New York: MacMillan Publishing Co.,

Krajewski, Lee C. and H. E. Thompson. 1981. Management Science: Quantitative Methods in

Context, New York: John Wiley & Sons,

Levin, R.I.., C.A. Kirkpatrich & D.S. Rubin. 1982. Quantitative Approaches to Management.
5th ed. McGraw-Hill, Inc.

Zamora, Elvira A. 2004. Basic Quantitative Methods for Business Decisions. UP College
of Business Administration.

***