See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4933120

A linear programming solution to the faculty assignment problem

Article  in  Socio-Economic Planning Sciences · February 1976

DOI: 10.1016/0038-0121(76)90008-2 · Source: RePEc

CITATIONS READS

51 2,664

1 author:

Jon Breslaw
Concordia University Montreal
29 PUBLICATIONS   234 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Jon Breslaw on 07 December 2017.

The user has requested enhancement of the downloaded file.

Socin-Econ. Ph. Ei.. Vol. i0, pp. ‘2
_ 7, ‘30. Pergamon Press 1976. Prmted in Grear Britain

A LINEAR PROGRAMMING SOLUTION TO

THE FACULTY ASSIGNMENT PROBLEM

JON A. BRESLAW
Department of Economics, Concordia University, Sir George WilliamsCampus, Montreal, Quebec, Canada

(Received 1t F~b~~ry 1976; revised I4 July 1976)

Abstract-This paper investigates the problem of assigning faculty to courses at a university. A program is developed
which is both efficient in that integer programming is not required, and effective, in that it facilitates interaction by
administration in determining the optimal solution. The results of some empirical tests are also reported.

I. I~RODU~ON gorithm required an integer programming solution; this

An annual problem for administration, be it at the high limits the efficacy of his model:
school, college or university level, is the assignment of
teachers to courses. The general form of the problem is to “the major barrier to immediate practical application of
find the solution to the following system: the model is the fact that in all but very small
(a) The determination of the set of courses to be offered, departments, the computer time for the execution of
and the number of sections in each. the zero-one integer programming algorithm was
(b) The assignment of teachers to courses. prohibitively large”.
(c) The assignment of courses and sections to rooms
and time slots, and the allocation of teachers to sections. This paper attempts to overcome this difficulty, and
(d) The assignment of individual students to courses to develop a model that has particular applicability to
and sections, based on their requests, in such a way as to universities. The particular scheduling needs of a univer-
yield a minimum of conflicts. sity are considered in Section II, and a model is developed
This system has to be solved simuItaneously to yield an in Section III. Empirical testing is carried out in Section
optimal solution. A number of school scheduIing packages IV.
and techniques have been developed[l, 21. Indeed the
solution to the Master Scheduling problem and the IL UNIVERSPTY SCHEDULING PROBLEM
student allocation problem (parts c and d of the general Universities tend to operate somewhat differently from
Sorm of the problem) is available as a commercial service high schools, in that the master scheduling problem does
for schools and junior colleges; allocations satisfying 99% not exist in the same way. Each department decides which
of student requests are regularty achieved even for large courses it will offer, how many sections of each and when
schools [3]. these sections will be offered. Effectively, this year’s
However, the allocation of courses to teachers, and the program tends to resemble last year’s, both in the number
determination of how many sections of each course are and type of courses and the times of the sections. This
given to a particular teacher (part b of the general fairly rigid structure ensures compatibility between
problem) is taken as exogenous to the master scheduling different departments. At registration, each student
problem. This allocation is usually done manually by the decides which course and sections he wishes to take;
department chairman and/or the school principal, in since section times are already available, he can select a
consultation with the respective teachers. A number of program which he knows is feasible, in that there are no
models have been developed to efficiently solve this time conflicts. Maximum section size is predetermined as
problem[2,4-6]. The state of the art is characterized by a matter of policy, so that when a section is full, it is
the work of Andrew and Collins[7]. An objective simply closed. If there is substantial excess demand for a
function, consisting of the weighted sum of teacher particular course, an additional section might be opened,
preference and instructor effectiveness (as evaluated by to be staffed by a part-time instructor; this procedure
the departmental chairman) is optimized, subject to occurs relatively infrequently.
course assignment constraints (every course must be The real problem is the allocation of faculty to courses
taught) and teaching load constraints (each instructor prior to registration. Since there is a fair degree of change
must teach a full work load). This is a standard in faculty preference from year to year, as many
transportation problem for which a number of efficient instructors will want to teach new courses, rather than get
algorithms exist, and which, by the nature of the model, stale, and given the normal staffing changes, visiting
guarantees integer solutions. appointments, leave of absence, etc. the problem is not
Tillett(61 expanded the model to include different trivial.
preference weightings by ins~uctors to allow for the A number of considerations thus have to be taken into
possibility that the preference of a teacher for a course account when constructing a model for use in the
could vary with the number of sections of that course university context.
assigned to that teacher. He also introduced a set of (1) The question of effectiveness is no longer valid in
constraints relating to the maximum number of new this situation-it is normally considered that if an
preparations acceptable to each teacher. Tillett’s al- instructor wishes to teach a course, then he will probably

SEPS VOL. IO NO. f-A 227

228 J. A. BRESLAW

be as effective in that course as in any other course that he course is totally taught by faculty j. Intermediate values
teaches. Presumably, in the extreme, a particular prefer- are unacceptable, since they would employ that some of
ence can be vetoed by the departmental chairman. the course is taught by one instructor, and part by
(2) Faculty preference must be taken into account, and another-an inadmissable situation. Hence Xii must take
the preference weightings must be a function of the the values of either zero or unity. pii is the preference
number of sections of a particular course that an rating for course i by faculty j.
instructor teaches. The university assignment problem can be formulated
(3) Since the timetable is already determined, a number as a linear programming problem as follows:
of constraints must be added to prevent possible conflicts.
(4) The usual teaching load and section assignment Max 2 C P+%
constraints must be met.
(5) One would like to apply such a package to a large subject to
department; this effectively prevents the use of zero-one
integer programming because of the prohibitive cost. gxi=l i= 1,2 ,...n (1)
These are the purely technical considerations. How- i=l
ever, past experience has shown that the application of
such an O.R. package does not satisfy administration. $Xij= 6 j=1,2,...n
After having made a trial run, one learns that the i=L
“optimal” solution is “not satisfactory”; it is politically j =1,2,...n
2X+1
unacceptable, socially unjust, or a number of other iEt k = 1,2,. . K
euphemisms for “it won’t do”. Some constraints, one
learns, have been omitted; some bounds are not as rigid as cxijsl j= 1,2 ,...m (4)
they were first presented. This sort of experience is not ieg
g=l,2,...G
new-Stimson and Thompson [8] clearly appreciated it:
i=l,2 n
xij30
“Thus, at the end of it all, the operation researcher j=l,2::::m.
has the optimal solution to a problem in which the
parameters, constraints and so forth are not optimally The objective function is the weighted sum of the X,
determined, and where his assumptions are not for each course and faculty, where the weights are
necessarily realistic in terms of the problems facing the derived from the faculty preference ordering (for
administration.. . Hence the solution to a linear methodology, see below). The first constraint set (1)
programming model.. should not be viewed as the requires that each courset is taught once, i.e. that the total
solution to the problem. Rather it should be judged for assignment value associated with each course must equal
its usefulness in a specific solution. unity. Constraint set (2) requires that each instructor
teaches his course load (Wj)-thus the total assignment
The package described in Section III solves the value for the jth instructor is lI$ Since it is normally
technical problem of a university scheduling problem but required that faculty split their work load equally between
does so in such a way that administration is actively semesters, this constraint will apply twice for each
involved in the problem, and hence effectively aids instructor, where W, represents the respective semester
administration in solving a real world problem. work load. The classical assignment problem consists of
these two constraint sets, together with non-negativity
IU. MODEL METHODOLOGY conditions.
The academic assignment problem is a variant of the Constraint set (3) stipulates that an instructor may not
usual assignment problem, which itself falls within the set teach more than one course in any one time slot (k); this
of problems categorized by the transportation problem. constraint holds for all instructors and all time slots. The
The transportation problem is profusely described in the inequality exists since it is permitted for an instructor to
literature[9]; the assignment problem is similarly well teach no courses in a particular time slot. In practice, this
discussed[lO, 111. The problem consists basically of constraint need only be considered for those time slots
minimizing a costs matrix (or maximizing a benefits where there exists a possibility of more than one course
matrix) that results from any assignment choice, subject being taught by a particular faculty member.
to the available staff and ensuring that all the jobs are Constraint set (4) permits an instructor to teach a
filled. maximum of one course from a particular cluster (g).
This problem could be solved using zero-one integer Again the inequality is permitted, since no courses may, in
programming; however, as discussed above, this becomes fact, be selected from a particular cluster. Finally,
infeasible for large problems. In fact, this is not constraint set (5) imposes non-negativity conditions on
necessary, since the ordinary SIMPLEX algorithm will the Xii.
give integer solutions to the transportation problem since Faculty list the courses they wish to teach in order of
the structural form of the constraint set guarantees that declining preference. If an instructor wishes to teach one
the extreme points of the feasible region will be integer. section of a particular course (or indeed one course out of
Consider the academic assignment problem. Let the a number of courses), the set of sections (or courses)
variable Xii be the percentage of course i taught by concerned is called a cluster. All the entities within a
faculty j. A zero value implies that faculty j does not cluster naturally have equal preference weighting. Should
teach course i at all, while a value of unity implies that the an instructor wish to teach another section of the same
course, but wishes this section to be assigned a lower
:The concept “course” will be taken to embody both the preference value than the first section, then a cluster of
concept course and section-i.e. a particular time slot in which a desired sections is given lower down in the preference list.
particular study is taught. Half year courses received half weights.
 A linear programming solution to the faculty assignment problem 229

The Simplex algorithm was used to solve the problem, B. A theoretical problem
and, as mentioned before, provides an integer solution. The following theoretical problem was set up: forty
The addition of inequality type constraints (3) and (4) does courses are to be taught in the first semester, forty in the
not change this result (see Appendix). In this case, since second and another forty are full year courses. Twenty-
we are dealing with a sparse matrix, the Revised Simplex six faculty each have a teaching load of 3 courses per
algorithm was used. Thus a global optimum is guaranteed. semester, and an additional instructor has a load of 2
This takes care of the technical side. The other courses per semester. Thus total course load and total
requirement was an interactive capability for administra- teaching load are equal in each semester. There are 20
tion. This was achieved by programming a complete time slots in each semester.
package, which carried out initial processing, linear To ensure a feasible solution a particular solution was
programming and final processing, and which prints output chosen. Each course was then assigned a time slot
at each stage. The initial processing produced a listing of randomly (based on a table of random numbers), with the
timeslotsandfacultyteachingloads;alistingof preassigned exception that if a course in the particular solution was
courses by faculty (a preassigned course is one for which assigned a time slot which involved a conflict with a
only one faculty has expressed a preference) for each course that was also in the particular solution, then a
semester; initial consistency checking; a listing of the new time slot was selected. Each faculty was assigned a
remaining unassigned teaching load for each faculty by total of nine courses-the particular solution, plus
semester; a cross-listing of unassigned courses, their time remaining courses, also chosen randomly. The order of
slots and feasible instructors; and a timetable showing all the nine courses was also randomly determined; the order
courses and the respective instructor for the preassigned was presumed to signify the preference ordering.
courses. Error messages indicated where no feasible An optimal solution was obtained in 13 seconds
solution was possible at any stage of the process. (decimal) of CP time (201 iterations). The optimal solution
This initial processing, before a linear solution was differed from the particular solution, and obviously
attempted, not only allowed interaction between the O.R. produced a higher valued objective function. For n = 120
staff and administration, but also saved a lot of computer (courses) and m = 54 (27 faculty, 2 semesters), this
time in attempting to solve problems for which no feasible implies a permutation matrix of 6480 elements, which is
solution existed. After a successful run, a listing of by no means trivial.
courses by faculty and semester, a full timetable and a
normalized table of preference values for each faculty V. CONCLUSION
were produced.
The results discussed here suggest that there are
considerable benefits to be gained by using optimization
IV. EMPIRICAL TESTING methods in the solution of the university scheduling
The efficacy of a particular program in solving problem. It is perceived, however, that these gains do not
assignment problems is di~eult to measure, since each accrue solely by utilizing an efficient algorithm, necessary
particular assignment problem differs from another, and though such an algorithm may be. Rather, these benefits
so no standard exists. Two approaches were taken: a real come about to a large extent by the bringing in of
world problem and a theoretical problem. Although administration during the O.R. process, rather than, as is
neither of these approaches is ideal individually, the the norm, of communicating before the process, and again
experience gained from solving these two problems, taken afterwards. It appears that this interaction can only be
together, gives some idea as to the usefulness of this type achieved if the output is presented in such a way as to be
of approach. precise, informative, and above all, familiar.
O.R. techniques have appeared to be most effective in
A. A real world problem dealing with small specific micro models, where the
system is well defined, and optimization takes place
The department of economics at Concordia University
(SGW), Montreal, has the usual timetable problem; within this specific system. At the macro stage, where the
manual solution, based on previous year’s experience, system itself is not even a given, O.R. performance is
averages three months. The department is large enough to dubious, to say the least. One can optimize a battlefield
provide a reasonable test. Each faculty member was deployment but that hardly leads to the desired result if, in
asked to give an ordered preference listing of the courses reality, one shouldn’t be fighting the war. if there is a
that he wanted to teach, of total length of twice his qualitative nature to the world, be it the preservation of a
species or the running of a school system, and if this
required teaching load-this provided sufficient slack,
qualitative nature is best appreciated by an administrator
hopefully, for a feasible set to exist. The problem covered
because perhaps of familiarity, or intuition, then it may
85 courses, some of which were one semester, and the rest
full year, 20 faculty, 20 time slots and 2 semesters. This well be that an interactive symbiosis of O.R. techniques
coupled with administration “input” may yield better
appeared to be a relatively easy problem, at least in
results than either one could do individually.
computer terms-an optimal solution was obtained in less
than 6 seconds of CP time (CDC 6400). It was clear,
however, that the most valuable output of the package Acknowledgements-The author gratefully acknowledges the
were the listings produced in the initial processing prior to encouragement and help he received from Morton Stelcner and
the linear programming solution. By giving administration Muriel Armstrong, both of Concordia University.
information in a recognizable and useful form, that
showed each stage of processing, a degree of communica- REFERENCES
tion developed that allowed far more interaction than is I.R. E. Holz, Computer-assisted scheduling. J. Ed. Data
usually the case. This is borne out by the fact that by using Processing 1, 36.41 (1964).
these techniques, the timetable task was finished in two 2. J. F. McNamara, Mathematical programming models in
weeks-a distinct improvement. educational planning. Rev. Ed. Res. 41(S), 419-446 (1973).
230 J. A. BRESLAW

E. Harper (Programming consultant, Cogito, Inc.), Private Consider a (one semester, for simplicity) problem, with n
communication (Nov. 1975). courses, m faculty, and r other constraints (inequalities). There
G. M. Andrew, Operations Analysis. University of Minnesota are n X m variables (Xl,), n + m artificial variables and r slack
Press (1970). variables. An initial feasible solution will eliminate all but one of
K. A. Fox and J. K. Sengupta, The specification of the artificial variables from the basis, since the equality constraint
econometric models for planning educational systems. KY/&S set consists of only n + m - 1 independent equations-sum of
21, 665-693 (1968). course load equals sum of teaching load. The n + m + r equations
6. P. I. Tillett, An operations research approach to the form a basis, whose solution is composed of at the most of
assignment of teachers to courses. Sot.-Econ. Plan. Sci. 9(3), n + m t r non-zero components. If r of these components solve
101-104 (1975). for the slack variables (i.e. all the slack variables belong to the
1. G. M. Andrew and R. Collins, Matching faculty to courses. basis), and one component solves for the remaining artificial
College University 46, 83-89 (1971). variable, there will be left at the most n t m - 1 non zero
8. D. H. Stimson and R. P. Thomoson, Linear proarammina. components for the matrix X. Since the permutation matrix is
busing, and educational administration. Sot.-E~on.P[on. SA. m x n, there exists a row or column with only one non-zero
S(4), 195-205 (1974). element in it. Since the constraints are all integer, the element
9. G. B. Dantzig, Linear Programming and Extensions. Prince- must also be integer. By removing that row (say), and changing the
ton University Press, Princeton, N.J. (1963). respective teaching load (W;), a new extreme point is reached
10. P. S. Dwyer, Solution to the personnel classification problem involving n - 1 rows and at the most m columns. Repeating the
with the method of optimal regions. Psychomefrika 19,l l-26 argument, it is clear that, providing the slack variables enter the
(1954). basis, that all the extreme points are integer.
11. J. von Neumann, A certain zero-sum two person game Initially, when the artificial variables all enter the basis, all the
equivalent to the optimal assignment problem. In Contribu- slack variables will have non-zero values, and hence also belong to
tions to the Theory of Games II (Edited by H. W. Kuhn and A. the basis. Consider the case of the ith variable entering the basis,
W. Tucker), pp.5-12. Princeton University Press, Princeton, such that the kth slack variable changes from a non-zero value to
N.J. (1953). zero (the inequality becomes an equality). However, when the ith
variable enters the basis, some other variable, say the jth, must
APPENDIX leave the basis, since the sum of the decision variables (i.e.
In the usual linear programming problem, the linear constraints non-slack variables) is constant; it equals the total course load.
define a convex polyhedron in an n-dimensional space. The Hence the slack variables always remain in the basis, since no
optimum point is found at a corner, or extreme point, of the decision variable can be altered unless some other decision
polyhedron. The optimal value of the objective function defines a variable is also altered. Consequently, if the slack variables
supporting hyperplane at this point. Even if we do not know the always remain in the basis, then, by the argument given above, the
form of the objective function, we know that if the constraints are extreme points are guaranteed to be integer, and hence the optimal
specified in such a way that the resulting polyhedron has integer solution must also be integer.
values for all its extreme points, then the solution to the L.P.
problem must also have an integer solution.

View publication stats