You are on page 1of 11

The Evolution of the Diet Model in Managing Food Systems Author(s): Lilly M.

Lancaster Reviewed work(s): Source: Interfaces, Vol. 22, No. 5 (Sep. - Oct., 1992), pp. 59-68 Published by: INFORMS Stable URL: http://www.jstor.org/stable/25061663 . Accessed: 21/11/2011 08:14
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Interfaces.

http://www.jstor.org

The Evolution Food Systems


Lilly M. Lancaster

of the Diet Model

inManaging

School of Business Administration and Economics University of South Carolina at Spartanburg Spartanburg, South Carolina 29303

The

electronic '60s with

revolution

in the

in food-systems the formulation of the

management least-cost-meals

began mathe

matical programming model. This model served as the basis for the CAMP (Computer Assisted Menu Planning) software de
veloped erences for mainframe computers. Over the next two decades,

researchers

introduced

techniques

for incorporating

food pref

Miniand micro-computer soft planning. ware Since the early have used math '60s, managers emerged. to enhance cost and nutrition ematical control programming consumer and to increase satisfaction.
the best possible meals at cost is the stated the least possible Preparing for management objective of food-systems most feeding programs. Menu planning is a key component since mines food, equipment, quirements planning experts the menu deter re and personnel 1975]. The menu both nutrition (RDAs) identify minimum daily allowances intake levels for 29 nutrients. Preventive health care agencies recommend upper limits for intakes of fat, cholesterol, and sodium. To the MS/OR researcher, plan

into menu

[Kotshevar process deceives

to satisfy these recommenda ning menus a tions represents problem with more than 30 nutritional constraints and thousands of variables. Furthermore, nutrition cannot be

and the public

simple procedure. Nutrition Board's

to be a by appearing the Food and However, [1980] recommended

the only goal. The consumers' food prefer ences must be considered as well as the
PROGRAMMING?MATHEMATICAL INDUSTRIES?FOOD

Copyright ? 1992, The Institute of Management Sciences 0091-2102/92/2205/0059$01.25 This paper was refereed.

INTERFACES22: 5 September-October 1992 (pp. 59-68)

LANCASTER
impact of the menu upon and the budget. nology, The increased personnel, tech on behalf menus.
able for

of the consumers displays

The menu
selection,

in planning the items avail


meal

of computers availability and of affordable food nu computerized opportunities for developing mathematical models for menu planning. Models in the conceived as linear programming '40s by economists have evolved over three decades problems into applications software. The Foundations foundation for the application of in planning diets is techniques [1945] "The cost of subsistence." The MS/OR trient data bases enhances

course-by-course,

by-meal,

or

day-by-day.

process decision-making the allocation scheduling, on a discrete more time scale. than one

this Technically is called menu of menu If the menu items lists

tive. Otherwise menu

item per course, it is nonselective,

it is selec and the

the meals. Mate defines completely rial requirements is based on the planning menu schedule. The menu items are de one can de fined by recipes, from which rive the quantities to of ingredients needed a portion or multiple prepare portions. The first application of mathematical programming regarded menu scheduling

Stigler's that satisfy nutri Finding food mixtures is a problem of solving tional requirements a linear system of equations with more variables (nu (foods) then equations When Dantzig trients). [1963] developed for solving linear pro the simplex method

A final problem
artistic planning.
as a sequential, problem. Each
mal combination

relates to the
of menu

he used Stigler's data gramming problems, to illustrate the viability and efficacy of the technique. Since then, the linear program ming formulation lem has continued of the classic diet prob to evolve. Smith [1963]

aspects

multi-stage optimization determined the opti stage


of menu items for a meal.

that incorpo tried to produce solutions food mixtures rated more acceptable by and up constraints adding proportionality to increase food variety. His per bounds model did not contain rules for converting as purchased (AP) foods into edible por tions (EP) based on recipes. Least-Cost Menu Planning Models An increasing segment of the population through feeding programs such as hospitals, institutions, nursing schools, and prisons and other homes,
food-systems rants, mobile operations, meal programs, such as and restau congre

The data fied

structures

of this model of future

also clari

food-sys tems management software [Armstrong, Balintfy, and Sinha 1982]. The least-cost meal model is an extension of the classic diet model. nents that the compo are fixed portions of menu of meals items. Menu items are mixtures of foods It is assumed

the organization

obtains meals

at

defined

the variable in by recipes. While the classic diet model is the food, the vari able in this model is the menu item. Menu is a decision items course
under

planning ing menu presumed


menu items

problem of allocat to meals. The meals have structure


consideration

gate dining agers and nutrition

facilities

for the elderly. Man experts make decisions

and the set of


are as

INTERFACES22:5 60

THE DIET MODEL


signed to courses. Menu planning meal but extends a menu mathematical is seldom limited to one approach and the growing of computers prepared the way availability for full-scale and software development control

to a sequence of meals or cycle. Variety plays an important role. The model achieves variety by sepa

items rating identical and similar menu over time by a minimal separation period based on separation ratings collected from a target population. tion data, planners menu of earlier potatoes separa By introducing can restrict the entry of in the function items into the model ifmashed solutions. For example, are in the solution for the first

applications [Gelpi et al. field trials and refine 1972]. Following ments, [1969] introduced an ac Balintfy ronym ning for computer assisted menu plan (CAMP) and the CAMP software. as part of the IBM contributed pro library, has been used as a reference

menu

CAMP,

grams in the development of commercial systems. describes CAMP in detail. [1975] Balintfy Since CAMP main, remains the number
and

in the public do of applications, modifi


is unknown. Sev

The

solution

to the

least-cost

cations,

extensions

diet is the equivalent human dog biscuit.


meal

of the

eral applications

a separation of the menu rating cycle, from of three prevents mashed potatoes in a solution again until the fifth appearing meal of the menu cycle. The solution to the model
tion, course

implementation duced 90-day and reduced food costs by five percent. CAMP was also used in its original config uration United in at least 18 institutions States and Europe. and Immonen in the In Finland,

have been published. One in a state hospital pro cycle menus [McNabb 1971]

Tusmi-Nuri menus

minimizes
structure,

cost subject
and policy

to nutri
con

[1982] planned for government-supported schools that lowered fat content, satisfied other

straints. Multiple choice programming are solved sequentially problems by (0,1) for each time period integer programming of the menu schedule. Balintfy veloped algorithm to solve such problems. pivoting In early installations, the computer-gen erated menus provided cost savings. Balintfy found that the menus a heuristic [1964] de with block

nutrient

and reduced food requirements, cost by six to nine percent. CAMP was also to generate fiscal reports for food extended
cost accounting [Fromm, Moore, and

Hoover

on 1980] and to manage menus line [Hoover and Leonard 1982]. In the

a 10 to 30 percent and Nebel [1966] were similar in ac

hospitality industry, Wrisley [1982, 1983] com used the CAMP concept to develop and control sys puterized food-planning tems. Wrisley's system is used commer to forecast sales, control cially by managers and compute recipe and menu inventory,
costs.

to those meals planned by tra ceptability ditional methods. Also, the menu met all the specified nutrient constraints, whereas menus used meth planned by previously con ods failed to meet nutrient constraints sistently. The proven effectiveness of the

The CAMP success problems

system enjoyed considerable in the '60s and '70s despite the common to such MS/OR imple

September-October

1992

61

LANCASTER
mentations. the decision
ers were

Since

the computer dominated process, many decision mak


of the computer-gener

suspicious

into a sequence of meals. The CAMP ap in solution technique, proach, a multistage a formulation of menu spired single stage planning This was accomplished problems. nutrient requirements for the by setting entire planning horizon and by converting separation into upper bounds on This model produces serving frequencies. the least-cost frequencies of menu items ratings

ated menus. mathematical difficult

Lacking

formal

education

in

programming, they found it to believe that the computer could and of

generate nutritionally adequate menus often recalculated the nutrient content the menus

by hand. With cost minimiza often suffered budget cuts tion, managers

An

institution
menu

offering
cannot content

a
control

for scheduling. Although this approach has not been implemented, the fast solution and shadow price information were useful for research. It led to the development linear programming food price index 1970]. of a

selective

the nutrient

or the cost

of a meal chosen by any individual.


In addition, because and loss of personnel. relied too heavily on solutions frequently low-cost menu items, menus often lacked related variety. A final problem tistic aspects of menu planning:
vor, temperature, and texture

[Balintfy, Neter, and Wasserman Preference Maximization Models Experience generation with CAMP

led to a second

of menu-planning models that consumer food preferences. quantified Food preferences and the desire for variety have recognized. Siegel and studied the effect of monot Pilgrim [1958] Zellner ony on food acceptance. [1970] long been studied food acceptance frequency. Rolls in relationship et al. [1981] to

to the ar color, fla For


such as

combinations

for a meal
example,

or intrameal
a monotone

compatability.
colored meal

serving proved

baked

potatoes, yellow and cantaloupe may not be very squash, even though nutrition and cost appetizing
controlled. However, no universally ac

fish, mashed

that preference for a food item is is in serving frequency penalized when creased beyond a quantifiable point be cause of the "sensory specific satiety" of repetitive eating. Benson [1960] suggested a functional between relationship and serving frequency. food

are

rules exist for in ceptable programmable trameal compatability. schedule determines While the menu which menu which menu a menu item is served in which meal, a

preference Balintfy et al. (1974) performed de experiments to establish the nature of the func signed tional relations between the measure item and of for a menu the time

the frequencies with plan determines items are served during the entire models produce cycle. Mathematical

con plan in a single stage if the for the whole straints are defined cycle. are optimal in terms of the cy Such plans cle constraints but have to be scheduled

preference history of previous erence for a menu concave since

The pref consumption. item was identified as a

function of time nondecreasing last consumption (Figure 1). functions and exponential Hyperbolic

INTERFACES22:5 62

THE DIET MODEL \f(t)


numerical to plus scale ranging three from minus to most three

(least preferred pre ferred). Then for the same list of items, in dividuals were asked to specify how many days
of

should
items,

pass between
or a separation

repeat

servings
Us

those

rating.

function was ing this data, an objective to maximize formulated For preference. item combi each meal the most-preferred nations ence erence Paulus compete for solution with both prefer the pref and

computations including and separation ratings.

Balintfy

this approach, but [1986] described the application of the preference-time function to menu planning was bypassed
in the '70s. Instead, endeavors were made

Figure sents of

preference-time increase in preference a menu item while particular the is not reaches eaten for a

1: The

function for a

repre portion

to extend model

considering

the time since it was


item ence The

last consumed.
long time,

If the

the planning horizon of the to a longer time period. from a meal Sinha [1974] computed the preferences

adage fonder" applies

its asymptotic highest makes the heart "absence to our favorite

the prefer value "a." grow The

repetitive serving of items over a longer planning horizon and as a result de the preference-frequency function veloped regarding (Figure 2) and explored concepts gramming erence is expressed nonlinear pro (Figure 2). Here, pref in the function of serv

foods.

"a"

parameter is potentially different for each item and is estimated by preference rating
surveys. The concavity consecutive or similar of the preference-time for time "ideal" the

function
t*, between same item

implies

that there is a time interval,


consumptions items where the is maximum. also be estimated This

average preference can time interval from surveys.

ing frequencies during the length of the horizon. The preference-time planning function and this preference-frequency
function show a one-to-one correspon

directly

dence with most to represent the time de

estimated using al parameters identical procedures [Balintfy and

are equally

useful

Melachrinoudis

The estima of food preferences. pendence tion of the parameters of the preference the con time function (Figure 1) utilized cept of separation ratings, CAMP system. Preference menu item were as did the ratings for each from a tar also collected

1982]. The major differ ence is in the data-collection phase. In are stead of separation individuals ratings, consume interval like to they would frequently menu items during the fixed-time of one month.

asked how

and used in the process. For get population individuals were asked to state example, their preferences for menu items using a

For the preference-frequency function, a function is the simplest approxi quadratic mation. Sinha used a target consumer pop ulation's preference and frequency ratings

September-October

1992

63

LANCASTER

\9M

menu

is the sum of quadratic prefer functions. The objective is ence-frequency cycle


to maximize a consumer's total preference

subject constraints.

to nutrition, The

structural,

and budget

solution

for menu serving frequencies the menu cycle. In verification school food service, involving Rump, and Sinha

provides optimal items during experiments Balintfy, this ap

tan 1a

dou proach. They in which preference ble-blind experiment menus maximized and control menus planned
were

[1980] used a 12-week conducted

by school
in an

food-service
alternating

specialists
sequence.

served

&

Results mized

showed menus

that the preference-maxi increased student participa

Figure 2: The preference-frequency


represents by consuming the total

function

tion by almost 10 percent. Plate waste de creased by as much as 24 percent. Food costs decreased, and the nutritional ade quacy of the menu The control menus cycle was guaranteed.

accumulated preference a of a menu item x portion

times during a fixed time period T, where T is a long time interval such as 30 days. It is
assumed that the item is consumed

tively with
time T. For

t = T/x
example,

repeti

time separations
if a consumer

during

planned by specialists did not always meet nutrient standards. The data base and the quadratic pro gramming model were also used
models. models, the

in food
In food pref

to a survey that he likes question fresh five times month, per 30-day peaches are t = or six there time separa 30/5 five-day for peaches that month. tions Conse during

responds to consume

service-policy-evaluation erence-maximizing

budget
as

is not minimized
Consequently,

but set arbitrarily


the para

quently
optimum the slope

g(x)

x.f(T/x)

and x* = T/t*
The

is the
of

a constraint.

serving at x =

the preference-time 1. The value of x* preference-frequency

tangent frequency. to the asymptote 0 is equal of in Figure shown function is estimated surveys. from the

metric

change of the food budget produces a corresponding change in the preference maxima (Figure 3). The lowest point of this preference-efficiency the least-cost diet the least-cost human menu diet curve coincides with to (LCD). The solution is the equivalent of the

the parameters of a quadratic function: g(x) = ax preference-frequency ? a is the consumer's bx2 where preference item (same as the a assymptote for a menu to estimate in Figure 1), x is the serving frequency, and b is the satiety coefficient numerical (a pen for serving an item too frequently). ality Total preference for all items in a given

of dog biscuit. The combination items may not be desirable for con

sumption, but nutrition and cost are con trolled. By updating the food-cost data base at the time of food purchases, the de
cision maker can always serve the least

cost meals

to meet

nutrient

standards.

In

INTERFACES22:5 64

THE DIET MODEL


contrast, the highest point on the curve, the cost of the affluent diet (CAD), is at is no the budget constraint tained when

300

g c
fe. "5 ~o

200

longer binding. There is no rationale for a budget to exceed this level. For example, are high-cost caviar and fresh asparagus items for items but not high-preference such as primary school many populations, on these children. Spending more money high-cost
ence. budget

100

items will

not

increase

prefer

-100

Arguments diate budget

can be made levels. One

to set interme

tion sets the budget an additional where most


Figure resents 3: The the preference total preferences curve efficiency for all menu rep

policy considera level at the point budget unit buys the

items in the objective


ence-maximized where budget menu items nutrient

function

of a prefer
model are

in preference, the best buy diet (BBD). At this point, a line from the to the preference effi origin is tangential increase
ciency curve. This solution represents serv

menu-planning and nutrient

constraints

imposed. Any point under


efficient menu and since plan are available constraints.

the curve is an in
for preferred the same lowest budget nutri

more

ing the meals most pleasure

that offer

the consumers

the

The

for the money. Another approach sets the food budget at the "cost of decent subsistence" (CDS) [Balintfy solutions 1979b]. Quadratic programming dual variables represent produce

tionally feasible point on the curve is the "least cost diet" or LCD. The LCD is the equivalent of a human dog biscuit where the
consumer's food preferences the budget preference The budget are disregarded

and the objective


nomically. As the metrically, model increases. the

is to feed cheaply and eco


is increased maxima of para the to

utilities of the constraints. ing the marginal is defined as an If the calorie constraint the marginal utility of calories can equality, be either positive or negative depending upon the budget. This implies that the last bite of food consumed
needs can cause either

cost-of-decent-subsistence

corresponding is the label

point on the curve where


with words, respect when to calories the consumer

the marginal
is zero. takes In other the

utility
last bite

to meet
a pleasant

energy
or an un

of food to satisfy his hunger, he or she still finds the food pleasurable. Other points of in terest are the BBD (best buy diet) and the CAD (cost of affluent diet). The BBD is the diet that gives the highest preference per dol
lar. The sive as a diet represents on is spent money high-cost and fresh caviar asparagus, CAD where items, even exces such

pleasant sistence where

sensation.

The

cost of decent

sub level is

represents budget of the calories the marginal utility zero. This allows the computation of

a unique

equivalent frequencies profiles.

serving preference-maximized for different food-preference of the application econometric using

though

is not increasing. It implies that preference high-cost items do not necessarily buy more
pleasure for consumers.

Taj [1984] extended quadratic programming

September-October

1992

65

LANCASTER
methods to estimate utility function coeffi cients. Taj tested estimation of these coeffi cients with United States Department of data. He estimated a (USDA) Agriculture nutrient of the meals composition [Balintfy and Prekopa 1966]. Gue and Ligett [1966] applied this prin In the ciple to selective menu planning. item with

quadratic utility function for the linear de mand system it generated, determining a specialized coefficients using stepwise procedure. Balintfy and Taj [1987] devel oped USDA family food plans using this the same models, Taj approach. Using the marginal [1990] computed utility cost of nutrient constraints. marginal Refinements quencies for planning and

CAMP model, menu

they replaced the entity of a the entity of a choice choice prob group of items. They obtained abilities from choices offered in past verified Liggett, and Cain [1968] also that the heuristic algorithm used in the CAMP software is the most efficient Gue,

menus.

in scheduling optimal fre resulted in a modeling concept

for this class of problems. The Gue and Ligett approach effect of time upon preference fect of preference upon choices. formulated tive menu a stochastic

ignores the and the ef Ho [1978] to selec

and scheduling. First, entr?e items were scheduled with maximum sepa courses were filled rations; the remaining items that were the entr?e least incompatible [Balintfy et al. 1978]. The

approach

with with

groups considered

the choice planning, composing that using mathematical techniques probabilistic preferences. the preference-probability He vectors

state hospital system of New Jersey ap this concept in a system that included plied a centrally controlled, locally optimized
computerized food-management informa

updated using a Markov sequential maximizing


model.

process prior solutions with a preference choice programming multiple

to deriving

tion system. It saved 10 percent of the food cost per person per day during the first year of operation while guaranteeing that nutrition were satisfied Menu standards [Balintfy for the menus 1979a]. so far are lim menus. Peo

Either pected
average

approach controls only the ex for the value of nutrients available


consumer. Proponents of selectiv

ity argue
nutrient

that it is unnecessary
constraints on either a

to satisfy
meal-by

Selective All

the models

Planning discussed nonselective

meal
tional

or day-by-day
allowances.

basis
However,

to meet

nutri
or a

ited to planning

if a meal

ple usually prefer selective menus, where they can choose among several items for
each course. However, an institution offer

fail to meet day's meals nutrient excesses ments, must

nutrient

require or deficiencies in subse

be carried over and offset

cannot control either ing a selective menu, the nutrient content or the cost of a meal chosen by any individual can overcome institution consumer. some The loss of con the It

quent meals planned by the computer. is to solution to this problem One possible offer meals that meet specified nutritional Using a computer and sequence meal selections plan, store, for nutrition and cost may be controlled next step in applying mathematical the constraints as choices. to

trol of nutrition

and cost if it knows

of the items being probabilities can then compute the expected

chosen. cost and

INTERFACES22:5 66

THE DIET MODEL


to food-systems programming ment [Lancaster 1987]. Summary
Food-systems management is a challeng

manage

References
Armstrong, 1982, erized R. D.; "The Balintfy, J. L.; and Sinha, of P. conceptual information management tems," Journal of Foodservice Systems, No. 1, pp. 47-58. food Balintfy, J. L. 1964, "Menu Communications foundation comput sys Vol. 2,

and mathe ing area for computerization In 30 years of expe matical programming. rience, using mathematical programming models cisions ally, to plan menus and make has proven cost effective. have reduced 10 percent. policy Gener de

these models

food costs

puter," No. 4, pp. 255-259. as J. L. 1969, computer Balintfy, "Systems/360 sisted menu Contributed planning," Program No. IBM Corporation, 260D-15.2.013, Library New York. Hawthorne, Balintfy, ming tions," Balintfy, agement, J. L. system 1975, for "A mathematical food management Vol. 6, No. 1, pp. "Computerized and decision program applica 13-31. food man

com by planning Vol. 7, of the ACM,

by approximately

ematical-programming planners have provided nutritional ment standards

Using math approaches, menu menus that meet

set by the govern

Interfaces, J. L. 1979a,

information

and other feeding programs. Using menu planners have traditional methods, fallen short of these standards and con tinue to do so. As models that include
consumer

systems,"

Proceedings of the 35th Conference of the Soci ety for theAdvancement of Food Service Re
search," 1979, Balintfy, pp. J. L. Richmond, 81-88. 1979b, "The cost of decent Vol. 25, No. subsis 10, Virginia, November 1-3,

have

been

con

structed
erences,

consumer

food pref
has in

tence," pp.

satisfaction

Management 980-989.

Science,

creased. Mathematical given institutions

budgets. On in setting the costs used a similar approach for family food plans. These food plans serve as the basis for the food stamp allot ments given
in

programming a basis for setting food a larger scale, the USDA has

has

Balintfy, J. L.; Duffy, W.; and Sinha, P. 1974,


food "Modeling erations Research, 727. Balintfy, J. L. and Melachrinoudis, E. S. 1982, preferences Vol. 22, No. over time," 711? Op 5, pp.

in the theory relations of repeti "Equivalence tive consumption," Naval Research Logistics Vol. 29, No. 4, pp. 651-657. Quarterly, Balintfy, ments ning," J. L. and Nebel, with computer Vol. Hospitals, J. L.; Neter, "An E. C. assisted 40, No. 1966, menu 11, pp. "Experi plan 88-96. W. between

to individuals

and families. pro

The application
gramming

of mathematical
management

food-systems

Balintfy, 1970,

J.; and Wasserman, comparison

has

fostered

the growth

of many

erized managerial
puterized systems,

techniques.
managers

comput Using com


standard

experimental

fixed weight
indexes," price tical Association,

and linear programming

food

can

ize recipes, control ingredients, forecast ture food requirements, schedule meals, and purchase and control inventory. Be cause of such systems, managers have the philosophies reexamined and policies basic to the industry. Mathematical-pro gramming techniques permit the modern food-systems and effective manager management to develop policies.

fu

Balintfy, sensory cess

Statis Journal of the American Vol. 65, No. 329, pp. 49-60. K. 1986, J. L. and Paulus, "Modeling over Interna time," specific satiety Vol. 37, No. 2, pp. 81-92. A. "Nature 1966, Prekopa, variations in the nutrient composi Health 148-169. G. T.; Sinha, P.; and Zolt Services Research, Vol.

tional Journal of Food Technology and Food Pro


Balintfy, random tion 1, No. Balintfy, ners, system Engineering, J. L. and of meals," 2, pp. A. of

efficient

J. L.; Ross, 1978, for

"A mathematical

preference-maximized

programming nonselec

September-October

1992

67

LANCASTER
tive menu matical 76. Balintfy, 'The on the D.; J. L.; Rump, of preference effect consumption Food Service 48-53. of and Sinha, P. 1980, menus planning Programming, and Vol. scheduling/' 15, No. Mathe 1, pp. 63 computer system," University sachusetts. McNabb, menus 4, 16, pp. Rolls, 1981, iology M. graphies PhD diss., for a decision School optimization of Management, Mas Amherst,

of Massachusetts,

maximized school

E.

1971,

"90-day Hospitals,

nonselective Vol. 45, May

lunches," Review, Vol.

School No.

Research

by computer," 88-91. E.; Rowe,

1, pp.

B.; Rolls,

E.;

Balintfy, J. L. and Taj, S. 1987, A Utility Maximi zation-Based Decision Support System for USDA Family Food Plans, Applications ofManagement
Science Series, Vol. 5, Randell Schultz, pp. editor, 1-22. JAI Press, Benson, to Greenwich, 1960, "A consumer Vol. B. 1963, 13, No. Linear Connecticut, psychometric preference," 1, pp. 71-80. Programming Press,

specific "Sensory and Behavior, Vol. on

and Sweeney, K. in man," satiety Phys 27, No. 1, p. 137. of

Siegel, P. S. and Pilgrim, F. J. 1958, "The effect


of monotony 756-759. Sinha, P. 1974, "Preference-maximizing PhD diss., models School of En Am the acceptance food,"

Americal Journal of Psychology, Vol. 71, pp.

P. H.

predicting

approach Personal

Psychology, G. Dantzig, tensions, ton, New

for menu and Ex Prince

planning,"

Princeton Jersey.

University

gineering, University herst, Massachusetts.

of Massachusetts,

Food and Nutrition Board, National Academy of Sciences 1980, Recommended Dietary Al
lowances. Fromm, B.; Moore, A.; and Hoover, fiscal L. 1980, for food

Smith, V. E. 1963, Electronic Computations of Human Diets, Michigan State University


Press, Stigler, East G. J. 1945, of Farm Michigan. Lansing, cost of "The Economy, Vol. subsistence," 27, No. 2, pp. of a

reports "Computer generated cost Di Journal accounting," of the American etetic Association, Vol. 170 77, No. 2, pp. 174. Gelpi, food M. J.; Balintfy, I. K. 1972, control Dietetic by J. L.; Dennis, "Integrated computer," Vol. L. C; nutrition Journal 61, No. and

Journal 303-314. Taj, S.

1984,

"Formulation

and

verification

quadratic programming model of the diet problem," PhD diss.. School of Engineering,
University sachusetts. Taj, S. 1990, of Massachusetts, Amherst, Mas

Findorff, cost

and of the 6,

American pp. Gue,

Association,

gramming

"Application in planning

of mathematical of human diets,"

pro

637-646. R. L. and

Methods
Vol. Tusmi-Nuri, foodservice

andModels

of Operations Research,
H. 1982, "School of Jour 1, pp. for ho Vol. 2,

"Mathemati J. C. 1966, Liggett, menu for hospital cal programming models Industrial Vol. 17, No. planning," Engineering, and for Cain, the K. C. zero-one

34, No.

5, pp. 395-410. S. and Immonen, in Finland. menu

II. Application planning Vol. system," 2, No.

8, pp. 395-400. R. L.; Liggett, Gue, J.C; of algorithms "Analysis ACM, Ho, P. gramming Vol. 1978, problem," 11, No. "A

1968, pro of the

computerized nal of Foodservice 1-6. Wrisley, tels," No. A. 1982,

Systems, "Food

Communications 12, pp. for 837-844. stochastic selective

cost

controls Systems,

ence-maximized

multi-stage model

prefer menus,"

Journal of Foodservice 1, pp. 59-69.

PhD diss., School of Engineering, University


of Massachusetts, Hoover, hospital L. and Amherst, Leonard, M. system Massachusetts. 1982, "Automated for di

Wrisley, A. 1983, "Food cost control for hotels Part II.A computerized food planning and
control tems, Vol. Zellner, G. system," 2, No. of Foodservice 2, pp. 95-125. vs. "Food 1970, acceptance Vol. March 44, Hospitals, Journal Sys

information

functions

serving 1, pp.

etetics," Journal of theAmerican Dietetic Asso


ciation, Kotshevar, National Vol. 80, No. 1975, 2, pp. 15-19. L. H. Institute Management by Menu, Indus for the Foodservice

frequency," 75-81.

try, Chicago, L. Lancaster,

Illinois. 1987, "Algorithmic interface with

INTERFACES22:5 68

You might also like