Quantitative Planning and Decision-Making Techniques (With Particular Reference To Agricultural Applications)
Quantitative Planning and Decision-Making Techniques (With Particular Reference To Agricultural Applications)
Quantitative Planning and Decision-Making Techniques (With Particular Reference To Agricultural Applications)
16
These developments, and many others, have ushered in the modern
quantitative era which is exemplified by the widespread application of
mathematics to planning problems of all descriptions
The majority of mankind's economic and industrial planning ideas are
quantifiable and therefore are susceptible to mathematical analysis The
rapid development of the electronic computer with its prodigious capacity
for calculation has also stimulated the application of mathematics to these
problem areas It is undoubtedly true that this trend will continue, and
probably accelerate, with the result that techniques which were formerly
thought of as pure mathematical abstractions will become the most modern
of planning methods This point is perhaps well illustrated by the present
application of Markov Cham theory to problems in agriculture and
industry
It is intended in this paper to describe the application of
(1) Mathematical Programming, and
(n) Markov Cham Theory
to some of the problems in agriculture, industry and economics with
which the Operational Research scientist is faced today
MATHEMATICAL PROGRAMMING AND PROBLEMS OF ALLOCATION
Allocation problems involve the allocation of resources to sectors of the
economy, to geographical areas, to industries or to individual jobs This
class of problem occurs when the available resources are m scarce supply
Therefore, it is desirable to allot the resources to the activities in such a
way as to minimise the total cost, maximise the total return, or attain
another such "rational" objective
Many allocation problems can be represented by a matrix as follows
Activities to be Accomplished
Resources
Ai
A2
An
Ri
Amount of
Resources
Available
R2
C21
C22
C2]
S:
RP
Cpl
Cp 2
Pn
mn
bm
Rm
PJ
mj
bl
Amount of
Resources
required
ax
a2
an
The entries in the cells, clj5 represent the cost (or return) that results from
allocating one unit of resource Ri to activity AJ5 and if the cost (or return)
17
from allocating an amount x^ of resource 1 to activity j is equal to x^ clj3
we have a Linear allocation problem
Allocation problems with linear return or cost functions have been
studied the most intensively because of the availability of the powerful
techniques of Linear Programming and also because of relatively small
data requirements
The application of these techniques to the problems of industry and
commerce involve the assumption that the amounts of resources available
(bi), the amounts required (aj) and the cost (c^) are known exactly This is
not always the case and it is sometimes desirable to test the sensitivity of
the solution to possible errors in these coefficients The technique known as
Parametric Linear programming allows one to carry out such sensitivity
analyses
LINEAR PROGRAMMING MODELS
A linear programming formulation can be written as either
(1) a primal model, or
(n) a dual model
depending on whether the objective is minimisation (primal) or maximisation (dual) Both types of model are connected by the duality theorems of
linear programming whereby the solution of one model contains the
solution of the other
The general statement of primal and dual can be expressed in matrix
algebra as follows
Mm
cX
Max
blZ
subject to restrictions A X ^ b
subject to A ^ ^ c 1
X^O
Z>0
where
c = l x n row vector
X = n x 1 column vector
A = m x n matrix
b = m x 1 column vector
Z = m X1 column vector
Note (I) The objective function and the constraints must obey the linearity rules
(n) A general method of solution known as the Simplex Method was developed in
the 1940s by G B Dantzmg, late of the Rand Organisation, and now a Professor
in the Computer Science Department of Stanford University
(in) Refinements to the general technique exist whereby it is possible to obtam(a) solutions to problems having integer and certain other non-linear constraints or costs, and
(b) parametric solutions
GEOMETRICAL STRUCTURE
18
The geometry of linear programming can best be illustrated by means of
the following simple two-variable model
max
3xx+4x2
subject to x 1 +2x 2 < 14
4x1+ :
Objective Function
Restriction I
Restriction II
Non negativity conditions
EXTREME
POINT
SOLN
Xi
19
sufficiently accurate mathematical representation of many physical,
economic and financial situations for it to be acknowledged as a major
planning and research tool It has been found particularly suitable for
application as a planning and research tool in the following fields
(1) Agriculture
E O Heady, W Chandler and others have pioneered, in the United
States, the application of linear programming methods to agricultural
planning both at the regional and state level They have used this technique
to indicate the optimum distribution of commodities subject to such
limitations as land, labour, capital etc In addition they have also used the
linear programming approach to investigate the effect of variable price,
yield and capital quantities on the optimum solution
In general the planning models usually take the form
Max Z=c 1 x 1 +c 2 x 2 +
+c n x n
Maximization of
"Profits"
subject to a n x 1 +a 1 2 x 2 +
+ainxn^b1
a2iX1+a22x2+
+a 2 n x n < b 2
Constraints due
to land, labour,
etc
!+a m 2 x 2 +
= 1,2,
+ amnXn < bm
n
land
capital
labour
equipment
etc
20
FIG
Month
December
January
February
250
240
200
30
20
20
220
220
180
March
April
230
255
20
60
210
195
May
June
310
320
10
20
300
310
July
August
320
300
10
30
310
270
September
October
November
240
260
260
90
60
50
150
200
210
Per monthly
group
620
405
600
580
560
21
Statistical and Social Inquiry
Quantitative Planning and Decision-Making
Techniques (with particular reference to Agricultural Applications)
TABLE 1
Constraints
Variables
(Crops)
Output/
Acre
Oats (X1)
Wheat (X2)
Barley (X3)
Corn (X4)
Apples (X5)
Seed
Potatoes (X6)
Main
Potatoes (X7)
Land
(Acres)
Capital/
Acre
('s)
1
1
1
1
1
98
12 06
10 2
88
10 4
06
08
06
08
03
03
04
03
04
10
06
08
06
08
27
208 5
61 57
40
11
10 9
140 39
90 70
40
11
10 9
66,280
1,316,329
92,338
32,856
199,214
31 54
30 27
35 45
37 55
10 6
Labour A Labour B
(M Days) (M Days)
Labour C
(M Days)
Designations
Requirements
j = l,
22
the acreages of corn, mam and seed potatoes in accordance with demand
and practicability Analysis of past records indicates that these acreages
could be restricted to
x 4 < 1,021
x 6 ^ 3,722
x7sc 14,253
The resulting solution to this augmented model yields a solution which
is more realistic in that the acreages for corn, seed and mam potatoes have
been constricted to withm acceptable limits of growth
Further levels of sophistication include the addition of constraints for
(l) A minimum level of acceptable income
(n) The allocation of available "tractor hours" to the various
enterprises
(in) The inclusion of various types of livestock such as cattle, sheep
etc (this, of course, requires the addition of new variables) etc
The ultimate aim is to produce a realistic agricultural planning model
for the region in order to examine m detail the effect of variations in
prices and quantities
2 Industrial Applications
The models which have been described in an agricultural context can
also be applied to many problems in modern industry The mathematical
programming approach has been applied to many diverse problem areas
which include
(l) Production scheduling
(H) Mixing problems m the chemical and feed-stock industries
(in) Problems of allocation
(IV) Transportation and routing problems
In each case the basic approach is to represent the situation in terms of a
mathematical model having an objective function which must be maximised
or minimised and subject to various constraints
The following simple illustration typifies the method of approach
TABLE 2
Turning
Boring
Profit per unit
x2
x3
6 mms
2 mins
2 mms
2 mins
4 mms
12 mms
12
24
Daily available
Production time
200 mins
160 mins
23
24
MARKOV PROCESSES - THEORY AND APPLICATIONS
Theory
A sequence of experiments that can be subjected to a probahstic
analysis is called a stochastic process If the set of possible outcomes is
finite then it is said that the process is finite
Stochastic processes can be classified by indicating special properties
by the outcome functions of the process
In a Markov Chain process there is a given set of states
and it is only possible for the process to be in one of these states at a given
time The process moves successively from one state to another and each
move is called a step The probability that the process moves from Si to Sj
depends only on the state Si that it occupied before the step The transition
probability Pu, describing the probability of moving from Si to Sj, is given
for every ordered pair of states In addition an initial starting state is
specified at which the process is assumed to begin
These characteristics can be readily expressed m a transition matrix P
Si
Sx
s2
sk
Pu
s2
sk
P12
P21 P22
Pik
P k i Pk2
Pkk
p2k
IfP=
05
05
0 25
0 25
0
0 25
0 25
05
05
P8=
04
04
04
02
02
02
04
04
04
25
AGRICULTURAL APPLICATIONS
An analysis of the size of dairy herds in Fermanagh for the period 1961/67
yields the following information
TABLE 1
No of Farms
Herd size
1961/62
1-9
10-19
20-29
30
3,003
635
84
23
1,126
705
69
23
3,745
1,923
Total
1967/68
26
TABLE 2
Herd
Size
0
1-9
10-19
20-29
30+
Total
1967/68
No of
farms
1961/62
(2,872)
3,003
635
84
23
2,772
1,700
138
61
23
3,745
4,694
1-9
10-19
20-29
30+
0
323
382
0
0
0
0
69
0
0
0
0
0
23
0
705
69
23
100
980
46
0
0
1,126
Note: (2,872) is the estimated number of potential farms that could enter into the
dairying field
0 965
0 566
0 217
0 726
1000
0
0
0 109
0
0
0
0 108
0 602
0
0
0 035
0 326
0 072
0
0
0
0
0
0 274
0
From the theory of Ergodic Marov Chains it is not difficult to demonstrate that P n tends towards
TABLE 4
0
0
0
0
0
936
936
936
936
936
0
0
0
0
0
050
050
050
050
050
0
0
0
0
0
013
013
013
013
013
0
0
0
0
0
001
001
001
001
001
0
0
0
0
0
It is clear that the number of potential entrants assumed affects the rate
at which changes in size take place and the rate of exit projected for the
industry
The effect of different numbers of potential entrants can be examined
through using the transition matrix and a device known as the T-vector
27
The T-vector indicates the proportion of the original farms plus the number
of potential entrants falling into each of the size classes, including O, when
sufficient time-periods have passed in order that the exit, gentry, growth and
decline are in equilibrium
Starting from the recognition that the T-vector for equilibrium is
obtained by solving
t P=t
n
given
it can be proved that
N
-(i)
N+C
and
N+C,
where n is the number of classes and N is the number of
potential entrants assumed
C and Cj are some constants which are independent of N
and are computed from the original probabilities
(I) N-2,875
Now
t=[0 936 0 050 0 013 0 001 0]
2,872
thus
to-O 9362,872+C
whence
C=196 38
1^=0 05 =
Q
Ci
whence
C!=153 42
t 2 =0 013
306C 38
2372+198 38
c2
3,068 34
whence
and
C 2 =39 88
C 3 =3 07
If M is the number of active farms in the initial time period, the number of
farms Fj m each category m equilibrium will be
(in) Fj =
Cj
(M+N)j j = 1, 2
N+C
In addition if E represents the net change in the number of active farms m
the dairying industry, then the net number of entrants or exits from this
28
industry is
N (M4-N)
-N
N+C
N(M-C)
whence
E=
-(iv)
N+C
Thus using formula (m) we can compute the number of farms in each size
class
153 42 [6,617]
3,068 38
= 3 3 0 85
say 3 3 1 0
[M+N]
N+C
2,872x3,548 62
=
N+C
3,068 38
3,321 5
say 3,321 0
Thus assuming a potential entry of 2,872 and also a continuation of the
trends indicated m the period 1961/62 - 1967/68, the following conclusions
can be reached at equilibrium
TABLE 5
Numbei r of farms
Reduc;tions
Herd
Size
1961/62
Equilibrium
No
1-9
10-19
20-29
30+
3,003
635
84
23
331
86
7
0
2,672
549
77
23
88 97
86 45
91 66
100 00
Exits
3,321
Total
3,745
3,745
3,321
/o
29
TABLE 6
Herd
Size
N=Number of potential
entrants assumed
2,000
2,872
4,000
402
104
8
0
331
86
7
0
283
74
6
0
Exits
3,231
3,321
3,382
Total
3,745
3,745
3,745
1-9
10-19
20-29
30+
CONCLUSION