CHAPTER 14

DECISION TREES
14.1 INTRODUCTION:

.
.

Graphical representation of the decision process in tenned as a decision tree. The decision tree shows details regarding the various decision alternatives, states of nature and their probabilities, and benefits/losses due to a decision.

.
.

The decision tree has a network of nodes and branches. There are two types of nodes namely decision node and state of nature (chance) node. The decision node is represented by a square and state of nature node is represented by a circle.

. .

There are two types of branches namely main branches or decision branches and subbranches or chance branches. The decision branches start from the decision nodes and represent the alternate courses of action or strategies. At the end of every decision branch there is a state of nature node
from which chance-branches emerge representing chance events.

Along these branches

are shown the payoff s and probabilities

of the respective

alternative courses and chance events.

The expected values of the outcome are shown at the tenninal end of the chance branches.
tI

14.2

FOLD BACK OR ROLL BACK PROCESS:

.
.

The expected monetary value of each chance node is calculated from backward of the decision tree (from right to left). Then the branch of a decision node connecting a chance node with the highest expected monetary value is selected. As we proceed backwards, the process is tenned as fold back
or roll back process.

14.2

OPERAnONS RESEARCH

EXAMPLE: The following profit data is available for a conventional lathe.

5000 2500

Show the above data on a decision tree.

SOLUTION:

,~

RS.5000

Rs.2,500
Rs.13,000

:'

Rs.- 3,000

Let EMV imply expected monetary vdue. (EMV of a decision alternative) = L (payoff of each state) . (Probability)

That is, [EMV of the decision alternative M{ie node 2}] = [5000 X 0.7] + (2500 X 0.5]

= Rs.4750

EMV of the decisionalternative

N (ie node 3)

= (13,000 x 0.7) + (-3000xO.5)

= Rs.7,600

14.3
I

DECISION TREE

.-

14.3 PROCEDURE

IN DECISION

TREE ANALYSIS:

(a) Detennine the decision points and the alternative courses of action at each decision point. I (b) For each decision point find the payoff and probability of each course of action. I (c) Starting ITomthe right end of the decision tree, calculate the expected monetary value (EMV) or expected payoff for each course of action. .

(d) Select the course of action which gives the best expected payoff for each of the decisions. (e) Proceed backwards to the next stage of decision points. (f) Keep repeating the above steps until the first decision point is reached. (g) As the last step, the course of action to be used ITomthe beginning to the end under , different possible outcomes for the whole situation is identified. 14.4 ADVANTAGES OF DECISION TREE APPROACH:

A decision tree shows the inter-related, sequential and multi-dimensional aspects of the decision making process.

.
.

The decision-maker will have all the possible outcomes, even if he is not going to use all of them. The decision making process can be easily and clearly understood by others. The assumptions made about the future is clearly reflected. It is very useful when an initial decisions outcomes will affect its succeeding decisions.

. . .
. .

14.5 LIMITATIONS OF DECISION TREE APPROACH: When there are a number of decision alternatives, the decision tree diagram very complicated. The decision tree becomes more complicated if more variables are introduced. The decision tree diagram becomes even more complicated when interdependent alternatives and variables are present. Time value of money is neglected. All calculations are averages.
It is difficult to assign probabilities.

. .

c.

14.4

OPERA nONS

RESEARCH

14.6 PROBLEMS ON DECISION TREES: PROBLEM: A customer asks an agent to sell three of his machines L, M and N and agrees to pay him 4% commission on each sale. He puts the following conditions.

.
.

The agent must sell the first machine 'L' within 45 days. When the agent sells the first machine 'L', he will be given the 4% commission on that sale. After this he can either backout or try to sell anyone of the other two remaining machines within 45 days.
-

I I I

I
1

I .
I

If he fails to sell anyone of the machines M or N within 45 days, he will loose the opportunity to sell the other machine also. But ifhe succeeds in selling the second machine (either M or N), he will be allowed to sell the third machine on the same condition mentioned above.

The following data has been given. (selling cost is the expense incurred by the agent when a scale is made).
MACHINE PRICE OF MACHINE (Rs) L M N 15,000 30,000 45,000 SELLING COSTS (Rs) 500 450 600

PROBABILITY OF MAKING A SALE 0.8 0.7 0.6

Draw a decision tree and determine the agent's best strategy.

SOLUTION:

The agent gets 4% commission ifhe sells a machine satisfying the mentioned conditions. The commission he receives and the actual profit he makes is shown in table.
MACHINE COMMISSION RECEIVED BY AGENT L 15,000 x 0.04 = Rs, 600 SELLING COSTS 500 PROFIT MADE BY AGENT (Rs) 600 - 500 = Rs 100

M N

30,000 x 0.04 = Rs, 1200 45,000 x 0.04 = Rs, 1800

450 600

1200 - 450 = Rs 750 1800 - 600 = Rs 1200

DECISION TREE

14.5

The decision tree is shown.

~",. ~. C:J~'" ~~/

Yo~

i.~v..eS ~.

ACCEPTS L

-?~o

Now let calculate the expected monetary value (EMV) or expected payoff of the various nodes using the roll back or fold back process. EMV of node 4

= [0.7 x 750 + 0.3

x 0] = Rs 525.

EMV of node 3 = [0.6 x 1200 + 0.4 x 0] = Rs 720

EMV of node M = [0.7 (750 + 720) + 0.3 x 0] = Rs 1029 EMV of node N = [0.6 (1200 + 525) + 0.4 x 0] = Rs 1035 EMV of node 2 = Maximum of EMV (M, N)

= 1035
EMV of node L = [0.8 (100 + 1035) + 0.2 x 0] = Rs 908
EMV of node 1 = Rs 908

14.6

OPERAnONS RESEARCH

INFERENCE: The path of optimal strategy is shown in double lines. Optimum strategy for the agent is

.
. .
Rs 908.

First he should try to sell!-. Ifhe sells L, he should try to sell N. Ifhe sdls N, then he should try to sell M.

The optimum strategy should be used by the agent to make an optimal expected amount of

PROBLEM:

A textile fabric processing unit wants to decide on drilling a borewell in its

plant premises. In that locality, only 30% of the borewells drilled were successful at a depth of 250 feet. The persons of that locality got water at 290 feet but the probability of getting water at 290 feet was only 15%. The cost of drilling is Rs 65 per feet. The life of the borewell was 6 years. In the net present value terms, the concern is paying Rs 21.000 for 6 years to a private water supply agency for supplying water. The concern has to take anyone of the following three decisions namely (a) Should it drill upto 250 feet? (b) If at 250 feet, no water is found, should it drill upto 290 feet? (c) Should th~ concern buy water from the water supply agency? SOLUTION:
The decision tree diagram for the given problem is a~ ~hown.

DECISION TREE

14.7

(250 X 65) = RS. 16,250 ~~<:> "-~O "~\~ ,,~~V~~~ ,,0 v<? :'\
Q ~vv ~<c,<c,.A X '),.':I\:) .
~<c,-.;-

~t-

~
\'=>

~<:> ~O~ . ~~':I

(290x65) =Rs.18,850

=Rs.18,850
'1>~~7 .000

Rs.{18,850

+ 21,000)

Rs.{16,250 + 21,000) =Rs.37,250

The expected monetary value (EMV) calculations are shown (use roll back process)

EMV of node Y = Rs (0.15 x 18; 850 + 0.85 x 37,250) = Rs 36, 700 EMV of node 2 = Minimum ofRs (36, 700, Rs 37, 250) = Rs 36, 700 EMV of node X = Rs (0.30 x 16,250 + 0.70 x 36, 700)
= Rs 30, 565.

EMV of node 1 = Minimum of(Rs 30, 565, Rs 21, 000) = Rs 21, 000 From the calculations it is seen that the concern should buy water from the water
-

supply agency as this decision yields the optimal least cost.

A welding rod manufacturing company has three alternatives with respect

to production'namely (a) Produce commercially, (b) build test pilot plants and (c) stop the production of welding rods. Ifthe test pilot plants are build, they will be 0.7 and 0.3. If the test pilot plant yields

14.8

OPERATIONS RESEARCH

high, the commercial plant also will yield high with a probability of 0.65. If the test pilot plant yields low, the probability that the commercial plant will yield high is only 0.2. If directly producing commercially is undertaken, the chances of it yielding high is 0.58. What will be tbe optimum decision of the company if the cost of a test pilot plant is Rs 3, 25, 000 and the profits earned are Rs 1, 30, 10, 000 and - Rs 15, 00, 000 for high and low yield conditions respectively. SOLUTION: The decision tree diagram for the given problem is as shown.
Rs. 1,30, 10,000

Rs. 1,30, 10, 000

a~~
f:y~i:;-
'

vO 0<>' q,~.

~~ ;)v
"'-c,

- Rs. 15, DO, 000

Rs. 1,30, 10,000

- Rs.15, 00 ,000 -

RS.O Using fold back process, the expected monetary value (EMV) calculations are shown.

EMV of node Y = Rs (0.65 x 1,30, 10,000+ 0.35 x (-15, 00, 000) = Rs 79, 31, 500

DECISION TREE

14.9

EMV of node Z = Rs [0.2 xl, 30, 10,000 + 0.8 x (-15, 00, 000)] = Rs 14,02,,000

EMV of node 2 = Maximum of (79,31,500,0)

= Rs 79, 31, 500

EMV of node 3 = Maximum of(14, 02, 000, 0)

= Rs 14,02,000

EMV of node W = Rs[0.7 x 79, 31, 500 + 0.3 x 14,02,000]

= Rs 59, 72, 650

EMV of nodel if the

[ test pIlot plantls

..

bUllt]

= Rs59,72,650-Rs3,25,OOO
= Rs 56, 47,650.

EMV of node x = Rs [0.58 xl, 30, 10,000 + 0.42 (-15, 00, 000)]

= Rs 69, 15, 800.

EMV of node I if commercial production = Rs69,l5,800. [ is done directly ]

Comparing the EMV of node 1 (when test pilot plant is build) with EMV of node I (when direct commercial production is under taken), it is seen that it is advantageous to go in for direct commercial production. Hence the company should not build the test pilot plant.

14.10

OPERAnONS RESEARCH

14.7

REVIEW OUESTIONS (DECISION TREES)

1. What is a decision tree? Explain its use. 2. Explain rolling back process in decision trees. 3. List the procedure in decision tree analysis. 4. What are advantages and limitations in decision tree approach.

-.-I