SOUTHERN JOURNAL OF AGRICULTURAL ECONOMICS DECEMBER, 1989

XLAYER: AN EXPERT SYSTEM PROVIDING MANAGEMENT
ADVICE TO COMMERCIAL LAYER MANAGERS
Ed Schmisseur and John Pankratz

Abstract Expert or knowledge-based system pro-
XLAYER, an expert/knowledge-based mi- grams contain domain-specific knowledge and

crocomputer program, was designed and de- use complex inferential reasoning to reach
veloped to diagnose layer management prob- conclusions human experts would reach if faced
lems and recommend expert remedial man- with a comparable problem (Hart; Weissand
agement advice. The program also provokes Kulikowski). The basic knowledge contained
management action by calculating the eco- in these programs consists of rules, facts, re-
nomic loss attributed to major management lationships, reasons, and heuristics obtained
problems. It analyzes data generated by a from human experts who can solve problems
commercially marketed layer performance fi- in a particular domain of expertise. Expert
nancial microcomputer program and has dem- systems embody techniques for solving prob-
onstrated the ability to emulate poultry man- lems, manipulating stored knowledge, coping
agement experts in the diagnoses of 80 indi- with uncertain information, and explaining how
vidual layer management problems. The pro- an inference is proceeding or a conclusion has
gram provides scarce expert poultry manage- been reached (Simons).
ment advice to poultry layer managers regard- Expert or knowledge-ased system program
less of size and scale of operation. developments are a new area of applied poul-

try science research. The results of one effort
to apply expert system technology to poultry

Key words: expert system, knowledge-based layer management are reported here.
system, rule-based, economic
loss, layer management.l, lr m. PROBLEM-SOLVING FOUNDATIONS

Expert or knowledge-based computer pro- The practical and empirically verifiable di-
grams are becoming an important operations agnostic rules of poultry layer management
and management component of many Fortune as practiced by a cooperating poultry manage-
500 companies (Simons; Herrod) and are being ment consulting firml served as the problem-
actively explored to assist smaller businesses solving paradigm used in the XLAYER pro-
(Harmon et al.). They are beginning to emerge gram. These domain-dependent methods were
as a field of research, development, and use in developed by consulting firms and have proven
commercial agriculture (McKinion and useful over a period of years through practical
Lemmon; Michalski et al.; Boulanger; Roach consulting experience. Collectively, these diag-
et al.; Fermanian et al.; Spahr and Puckett; nostic rules implicitly seek to maximize layer
Levins and Varner; Oltjen et al.). They also flock profitability by identifying problems
represent the next logical step in the progres- which impact production performance, costs,
sion of computer-supported information and and returns. They utilize problem-solving tech-
decision-aid systems for farm managers first niques and data manipulations which explic-
initiated by agricultural researchers in the itly and rigorously critique these separate but
early 1950s. (Schmisseur and Doluschitz). interrelated areas of layer management.

I The management expertise contained in the XLAYER program represents that of Mel Gehman, President and Founder of Heritage
PMS, Inc., RD #3 Box 458-B, Annville, Pennsylvania 17003.

Ed Schmisseur is an Associate Professor, and John Pankratz is a Courtesy Appointment, Department of Agricultural and Resource
Economics, Oregon State University.

Approved for publication as Technical Paper No. 8907 of the Oregon Agricultural Experiment Station.
This project was financed in part by the Department of Agricultural and Resource Economics and the Oregon Agricultural Experi-

ment Station, Oregon State University and Heritage PMS, Inc., of Annville, PA.
Copyright 1989, Southern Agricultural Economics Association.

183



The general problem-solving approach em- financial performance, and egg gradeout. Flock
bodied in the XLAYER program is relatively performance data contain livability, egg pro-
straightforward. It attempts confirmation of duction, and nutrition information such as mor-
all management problems contained in the know- tality, body weight, house temperature, egg
ledge base by matching a list of management case weight, shell thickness, metabolizable
problem symptoms with indicators of substan- energy, and protein, and the content of spe-
dard flock performance. When a critical num- cific amino acids in the layer ration. Financial
ber of matches occurs, a management problem performance indicators report itemized income
is confirmed and a management recommen- and expense categories and net profit. Egg
dation is issued. The search for management gradeout data show egg size and grade, qual-
problems begins with the evaluation of pro- ity, quantity, and price information.
duction decisions followed by decisions about The XLAYER program uses approximately
egg marketing, and then production costs. 100 different data observations generated by

The diagnostic rules, general problem- the Layer Performance program. These data,
solving approach, and search strategy embod- automatically stored in a file called "Record,"
ied in the XLAYER program were solicited represent both actual layer performance and
from the cooperating consultants. A series of equivalent performance potentials or stan-
interviews with these experts dards. Performance potentials or standards are
1) identified important management problems based on standards published by commercial
impacting layer flock profitability; genetic companies and breeder farms. They
2) dictated the data and the heuristics needed reflect the different layer strains, age of birds,
to diagnose these problems; molting phase, and environmental conditions.
3) established the analytical or step-by-step Other data required by the XLAYER pro-
procedure followed to determine problems; and gram, but not produced by Layer Perform-
4) ascertained what management recommen- ance, include information on pullet flock his-
dation would be issued. tory, layer house equipment, marketing ar-

After each interview, new insights about the rangements and prices, and criteria by which
diagnostic process were encoded in the flock performance are compared to perform-
XLAYER program and then tested using test ance standards. Pullet history information in-
case data. In subsequent interviews, previ- cludes pullet weights, feed consumption, uni-
ously discovered problem-solving rules were formity and shank index information at se-
confirmed, new problem-solving capabilities lected weeks of rearing, and general pullet
were added, and the depth of the program's rearing conditions. This information is manu-
diagnostic capabilities was increased. ally entered and permanently stored in a file

called "Flock Profile" when a pullet flock starts
DATA REQUIREMENTS production.

The XLAYER program was developed to Information about layer house equipment
analyze weekly layer flock performance data includes the type of feeding, egg gathering,
generated by a commercially-marketed poul- and manure handling system in the layer
try management microcomputer program, house. These data, also manually entered, are
Layer Performance. 2 This integration was permanently stored in a file called "Housing/
important since the consulting firm developed Equipment Profile."
Layer Performance to collect and calculate Information about marketing and prices in-
data useful in their diagnoses of layer man- eludes egg selling/processing arrangements,
agement problems. Furthermore, many of egg prices by egg grade or case weight, and
XLAYER's data needs could be obtained di- feed ingredient prices. These data are speci-
rectly from data files produced by Layer Per- fled in a file called "Price Profile." Price infor-
formance rather than through additional key- mation is manually updated weekly.
board entries. The various criteria by which flock perform-

The major types of data produced by Layer ance are compared to flock performance stan-
Performance and utilized by the XLAYER dards are contained in a file called "Evalu-
program are illustrated in Figure 1. The Layer ative Criteria." These criteria were obtained
Performance program produces a report which from the cooperating consulting firm. They are
provides weekly data on three major areas of based on experience in analyzing flock per-
poultry management-flock performance, formance records.

2Layer Performance is a registered trademark of Heritage PMS, Inc., RD #3 Box 458-B, Annville, Pennsylvania 17003.

184



Layer
Report

Flock Egg
Performance Gradeout

Egg Income Profit Size/Price
Production Grade

Livability Nutrition Expenses Quality

Figure 1. Heritage's Layer Performance Report.



PROGRAM OVERVIEW infer management problems based on other
known information. At each query point, us-

The XLAYER program: ers can ask the XLAYER program to explain
1) identifies major management problems its reasons for the query. Also, at the end of

judged to be impacting layer flock profitability; the consultation, users can query the program
2) calculates associated economic losses at- to explain how it arrived at its conclusions.

tributed to major management problems; and After the XLAYER program has completed
3) issues specific management recommenda- its analysis, an "Executive Management Re-

tions designed to eliminate the management port" is compiled. This report includes a brief
problem. It is written using a commercially- description of each management problem, its
marketed expert development and delivery associated economic loss to provoke manage-
shell microcomputer program, M.1.3 The ment action and facilitate quick partial bud-
XLAYER program' s code was written in the geting of the decision recommendation, and
"PROLOG"-like production rule language con- specific management recommendations. If no
tained in the M.1 program. The shell program management problems exist, the XLAYER
requires an IBM-PC compatible microcom- program generates a brief report indicating
puter equipped with a graphics card and a this finding.
minimum of 320K RAM. 

The XLAYER program's relationship to its DIAGNOSTIC FRAMEWORK
data needs and users is illustrated in Figure
2. Users execute and communicate with the A generalization of XLAYER's diagnostic
XLAYER program by interacting with M.1. framework is presented in Figures 3 through
The M.1 program automatically calls the 5. These diagnostics are contained in some 400
XLAYER program and immediately begins individual production rules. Diagnostics per-
the consultation. During consultation, the tain to egg production, egg blend price, vari-
XLAYER program automatically seeks rele- able production costs, and fixed costs. The di-
vant flock performance, performance stan- agnostic framework, or fault tree, illustrates
dards, and other needed data. Should data not the complexity of the inference process. It also
be found, the natural language user interface shows the path of impact of management
of M.1 queries the user. Users then must re- decisions made in each specific area of pro-
spond. A response of "unknown" is acceptable duction.
as the XLAYER program has the ability to Management decisions about layer nutrition

Layer Performance Report

<-1 aRecords & Standards 

^ —Pro |Flock Profile 

U se r— Housing/Equipment Profile

M.1
M. 4- |Price Profile 

XLAYER[_ ItXLAYER IEvaluative Criteria 

A- I' Questions I . | Management Report

Figure 2. XLAYER's Relationship to Users and Data.

3 M.1 is a registered trademark of TEKNOWLEDGE, Inc., P.O. Box 10119, Palo Alto, California 94303.

186



depict one of the complex inferences within plementation problem can manifest itself in
the XLAYER program. Nutritional decisions several performance indicators.
impact egg production directly and also indi- When a methionine problem is confirmed,
rectly through feed consumption (Figure 3). another XLAYER rule is invoked which actu-
But, nutrition decisions also indirectly impact ally issues the management alert and pre-
egg blend price through both small eggs and scribes a remedy. This rule is:
egg quality (Figure 4) and indirectly impact
variable production cost through feed costs f methionine and display ([Verify suspected
(Figure 5). Specifically, egg quality is directly methionine imbalance in the layer ration.
affected by shell quality, which in turn is im- Methionine intake should be within 20 per-
pacted by nutrition decisions, and feed cost is cent of your flock's potential intake. If con-
directly affected by feed consumption, which firmed reformulate the rations protein and
in turn is impacted by nutrition. Similar com- methionine content.
plex inferences exist for management decisions then problem.
affecting environmental conditions in the layer However, before the primary rule can be
house, disease, pullet quality, and stress. confirmed, either low feed consumption and

The complexity of the diagnostics contained low production or small egg sizes must also be
in the XLAYER program is more easily seen confirmed from other production rules con-
in its diagnostic rules. For example, the pri- tained in the XLAYER program.
mary rule confirming a methionine supplemen- The low feed consumption, low production, and
tation problem exhibits a complex inference. small egg sizes diagnostic rules are relatively
This rule and others presented also provide simple. The low feed consumption rule is:
insights into the basic analytical process or
problem-solving approach employed in the iffeedperhenay = FPHD
XLAYER program. The rule, taken directly and goalfeedperhenday = GFPHD
from the XLAYER code, is: and critical_feed_perhen_day = CFPHD

and FPHD < GFPHD * CFPHD
if ((low_feed_consumption and low_production) then low_feed_consumption.

or (low_feed_consumption and small_egg_sizes
ois unkn( onsumption and smThis rule confirms low feed consumption whenis unknown)

and methionineperhenday = MPHD the amount of feed consumed per hen day is
and goalmethionineperhenday = GMPHD less than the goal or standard for feed con-
and uppertwidebounds p UW B sumption per hen day adjusted downward by
and louer wide bounds = LWB coefficient of tolerance for feed consumption.

and MPHD > GMPHD * UWB The low production rule is:
or MPHD < GMPHD * LWB) if eggs_per_hen_housed = EPHH
or (small_egg_sizes and goal_eggs_perhen_housed = GEPHH
and methionine_per_hen_day = MPHD and critical_eggs_per_hen_housed = CEPHH
and goal_methionine_per_hen_day = GMPHD and EPHH < GEPHH * CEPHH
and upper_moderate_bounds = UMB and egg_blend_price = EBP
and lower_moderate_bounds = LMB and starting_number_hens = SNH
and MPHD > GMPHD * UMB and (((GEPHH - EPHH) * SNH)/12) * EBP =
or MPHD < GMPHD * LMB) PRODL
then methionine. and display( ['

The rule's interpretation is: Economic losses attributed to low produc-The rule's interpretation is:The rule '£ mterpretat ' .11n tion are: ',tab(6), $(PRODL)] )1) if low feed consumption and low produc- i r 
tion exists and the methionine level fed per then lowproducton.
hen day exceeds either an upper or lower It confirms low production when egg produc-
bound, tion per hen housed is less than the standard
2) or small egg sizes exists without either for production adjusted downward by a coeffi-
low feed consumption or low production and cient of tolerance for egg production, and it
the methionine level fed per hen day ex- calculates the associated economic loss. This
ceeds either a narrower upper or lower loss is based on the difference between the
bound, egg production standard and actual egg pro-

then a methionine supplementation problem duction per hen housed adjusted by the total
is confirmed. It shows that a methionine sup- number of hens housed.

187



Egg
Production

Quality Conditions Consumption 

1 Disease Pullet Disease Fright
Nutrition Parasites Quality Parasites

Environmental Ration Mechanical Environmental Cannibalism
Conditions Formulation Equipment Conditions Prolapse

Figure 3. Egg Production Diagnostic Framework of XLAYER.



Egg Blend Price

Market Small
Practices Egg Quality Eggs

I

Shell Interior
Quality Quality

Equipment Stress Disease Candling Stress Off Taste
Damage Error Discoloring

mnvroma ImIn1ral [7Handling &
EnvironmentalNutrition Disease Equipment

Conditions Absorption Problems

Low Feed Parasites Pullet
Figure 4. Egg Blend Price Diagnostic Consumption Nutrition Diseases Quality

Framework of XLAYER.



High Variable
Cost

Medication High Energy Miscellaneous Labor Wage Labor Wage
Costs Costs Feed Costs Costs Rate Hours

Costs

Feed Ingredient
Consumption Costs i

Interest & RepairI | + | +LDepreciation Costs
Environmental Feather

Conditions Wasteage Loss

Nutritional Theft

Figure 5. Variable and Fixed Cost Diagnostic Framework of XLAYER.



The small egg sizes rule is: in new grains gradually even if the cost
if caseweight = CW per pound is higher. Gradually move to

and goal_caseweight = GCW the lower cost grain substitute.'] )
and critical_case_weight = CCW then problem.
and CW < GCW * CCW
and CaverageWblendrice = ABP Notedly absent from the diagnostic frame-
and dozen_eggs_graded = DEG work are interrelationships between such fac-
and potential blend_price = PBP tors as nutrition, disease, and stress. Although
and (PBP - ABP) * DEG = SMALLEGGS these interrelationships are theoretically pos-
and display( [' sible, the cooperating consultants did not re-

Economic loss attributed to small egg sizes quire analyzing these type of interrelationships
is approximately,',tab(4), $(SMALL in their consulting work and did not include
EGGS), per week.'] ) them in the XLAYER program. They argued

then small_egg_sizes. in today's skillfully managed commercial poul-
try flocks, management problems attributed

This rule confirms small egg sizes when egg to these type of interrelationships are rare.
case weight is less than a case weight stan-
dard adjusted downward by a tolerance level PROGRAM OUTPUTS
for egg case weight. It also calculates the as- T X p 
sociated economic loss due to small egg sizes . The XLAYER program possesses the abil
based on the difference between the egg blend t to dagnose ultiple management problems
price received and the potential egg blend andrecommendmanagementactions formore
price multiplied by the number of eggs graded than 80 individual layer production manage-

Another type of diagnostic rule contained in ment problems. This capability is exhibited in
the XLAYER program reflects a complex in- a management report displayed on screen and/the XLAYER program reflects a complex in- or printed. The report identifies management
ference requiring the user to respond to a prte. The reort idifis management
query about current production conditions. A problems, their associated economic loss, and
simple example of this type of rule is: specifc management recommendations. An

example management report appears in Fig-
if (low_feed_consumption and low_production) ure 6. Specific management problems and rec-

or (low_feed_consumption and ommendations related to housing and equip-
small_egg_sizes) ment management, nutrition, diseases, eco-

and grain_change = yes nomics, marketing practices, and general
then change_grain. management practices are contained in the

As this rule is being confirmed and the pre- program.
conditions of either low feed consumption and Approximately 37 percent of the XLAYER
low production or low feed consumption and program's diagnostics are about layer nutri-
small egg sizes exist, the user is automatically tion Nutritional diagnostics include such
queried about possible recent changes in the things as excessive or restrictive metabolizable
type of grain used in the layer ration. The energy, calcium, and sodium supplementation,
query involves two separate rule statements methionine, methionine-cystine, lysine, thre-

onine, and trytophane amino acid imbalance,
question(grain_change)=' improper vitamin and/or trace mineral mix

Have you significantly changed the type supplementation, and poor egg shell quality
of grain used in your layer ration?'. induced by improper ration formulation.

legalvals(grain_change) = [yes,no]. Some 23 percent of the XLAYER program's
The first of these two query rules issues a diagnostics directly address general manage-
prespecified question to the user, while the ment practices. Improperly trained egg can-
second provides acceptable answers and error dlers, a poorly managed poultry waste sys-
checking capability to the query. tem, cage overcrowding or under-utilization,

If the user responds to the questions by an- sudden changes in layer house environment,
swering yes, another XLAYER rule is invoked excessive egg handling time, and egg or feed
which states the management problem and of- theft represent some of the type of general
fers a remedy. This rule is: management problems considered by the

XLAYER program.
if changegrain and display( [' Housing/equipment diagnostics account for

A major change in the grain ration is about 18 percent of the XLAYER program's
suspected to be causing production prob- management expertise. The XLAYER pro-
lems. Reformulate your ration and phase gram identifies problems like improperly ad-

191



justed or operating mechanical egg gathering PROGRAM TESTING/VALIDATION
and feeder equipment, high or low layer house The XLAYER program has been subject to
temperature, low or high water consumption controlled testing with numerous test case
due to malfunctioning waterers or poor well data. During initial case testing, the XLAYER
water quality, limited lighting, and inadequate program's diagnostic capabilities were further
ventilation. veno dti crlastion. 1 p enhanced, and its depth of knowledge was in-

Economic diagnostics, representing 14 per- cresed. More than 300 individual cases, rep-
cent of the XLAYER program's capability, resenting a flock's complete weekly data set,
address traditional economic performance were subjected to analysis by the XLAYER
problems. These include such things as high program. These cases, although carefully con-
repair and maintenance, pullet rearing, labor, structed, confirmed that the program could

interest, and energy costs. successfully diagnose each, as well as various
Disease and marketing problems represent practical combinations, of the more than 80

the remaining diagnostic powers of the pro- layer management problems included in the
gram. High broker fees, inappropriate class XLAYER program
weight sales contract, respiratory diseases, Currently, it is being intensively field tested
parasites, infectious coryza, fowl cholera, and with flock data from both the East and West
MG disease are some of the disease and mar- coasts. In these tests weekly flock perform-
keting diagnostics addressed by the XLAYER ane data are being analyzed by the XLAYER
program. program and its diagnostics are compared to

PROGRAM LIMITATIONS Pthose independently developed by the poultry
management experts being emulated by the

The XLAYER program has the potential to XLAYER program. In these limited field tests
be a useful management tool, yet it has limita- with some 122,000 hens in various flock sizes,
tions. Major limitations relate to the program's representing approximately 70 flock weeks of
knowledge base. XLAYER's knowledge base production data, the XLAYER program has
is static since the program does not possess exhibited the promising ability to perform on
the ability to learn from experience. As a re- a level consistent with that of its counterpart
sult, XLAYER's knowledge base must be con- poultry management experts. That is, in ap-
tinually updated as new diagnostic techniques proximately 90 percent of the weekly tests,
are developed and possibly new genetic bird the XLAYER program's diagnostics were
strains are introduced. The magnitude of this identical to those independently prescribed by
limitation is unknown. It may be mitigated the poultry management experts.
because new knowledge can be added to ex- Additional field testing and more intensive
pert system programs much more easily than validation with more participating flocks is
modifying code in algorithmic programs. continuing. Field testing is scheduled to be

XLAYER's knowledge base is also limited completed by the summer of 1990.
to the diagnostic knowledge of the cooperat-
ing consultants. No measure of their poultry PROGRAM AVAILABILITY
management diagnostic skills has been made. The XLAYER program will be made avail-
The magnitude of this limitation is also un- able to poultry managers by the cooperating
known. consulting firm after extensive field testing is

Economic losses attributed to low production are: $425 per week.

Verify suspected high metabolizable energy level in layer ration. If confirmed, reformulate ration to
reduce metabolizable energy level to your flock's standard recommended in the Layer Performance
Financial Report.

Also note a methionine-cystine imbalance in the layer ration. Methionine-cystine should be within
20 percent of your flock's potential intake. If confirmed, reformulate the ration's protein content and
methionine-cystine supplementation.

Water intake appears high.

Figure 6. XLAYER's Management Report: An Example.

192



completed in the summer of 1990. The cost of gram indicates that it can substitute for scarce
the program has not been established. Pro- and relatively expensive human, poultry layer
gram code, excluding Layer Performance and management expertise. Its successful adoption
M.1, can be obtained by contacting the authors and widespread use should increase the pro-
of this article. duction efficiency of layer operations and im-

rCONCLUSIONS prove producer profits.
The XLAYER program also has shown that

The present level of performance of the it can provide valuable management informa-
XLAYER program suggests that expert sys- tion in a readily useable form to a range of
tem programs have potential applied applica- flock sizes. Relative to economies of size is-
tion in poultry layer management. Further- sues, this expert system program appears to
more, it has been demonstrated that an ex- be size neutral. That is, it equally benefits dif-
pert system program can be linked to a stand- ferent sizes of operation. If any unequal bene-
alone agricultural, computerized decision- fit should occur, it is hypothesized benefits
support program and can automatically and favor smaller operations because of the econo-
routinely access information generated by this mies associated with making available scarce
program to diagnose management problems and expensive human expertise in the more
and provide insightful management advice. economical form of an expert system micro-
The level of performance of the XLAYER pro- computer program.

REFERENCES
Boulanger, A.G. "The Expert System PLANT/cd: A Case Study in Applying the General Pur-

pose Inference System ADVICE to Predicting Black Cutworm Damage in Corn." M.S.
Thesis, Computer Sci. Dept., Univ. Ill., Champaign-Urbana, IL, 1983.

Fermanian, T.W., R.S. Michalski, and B. Katz. "An Expert System to Assist Turfgrass Manag-
ers in Weed Identification." In Proc. 1985 Summer Computer Conference. July 22-24,
Chicago, IL, American Society of Agricultural Engineers, St. Joseph, MI, 1985. 499-502.

Harmon, P., R. Maus, and W. Morrissey. Expert Systems Tools and Applications. New York:
John Wiley and Sons Inc., 1988.

Hart, A. Knowledge Acquisitionfor Expert Systems. New York: McGraw-Hill Book Co., 1986.
Herrod, R.A. "Industrial Applications of Artificial Intelligence." In Proc. 1987 IEEE Western

Conference on Computer Systems. June 2-4, Anaheim, CA, IEEE Computer Society Press,
Washington, DC, 1987. 196-202.

Levins, R.A., and M.A. Varner. "An Expert Diagnostic Aid for Reproduction Problems in Dairy
Cattle." Comput. Electron. in Agric., 2(1987):47-56.

McKinion, J.M., and H.E. Lemmon. "Expert Systems for Agriculture." Comput. Electron. in
Agric., 1(1985):31-40.

Michalski, R.S., J.H. Davis, V.S. Bisht, and J.B. Sinclair. "A Computer-based Advisory System
for the Diagnosis of Soybean Diseases in Illinois." Plant Diseases, 67(1983):459-63.

Oltjen, J.W., L.G. Burditt, and G.E. Selk. "Expert System for Culling Management of Beef
Cows." Okla. Agric. Exp. Stn. Rpt., MP-119, 1987.85-88.

Roach, J.W., R.S. Virkar, M.J. Weaver, and C.R. Drake. "POMME: A Computer-based Consul-
tation System for Applied Orchard Management Using PROLOG." Expert Syst.,
2(1985):56-69.

Schmisseur, E., and R. Doluschitz. "Expert System Insights: Future Decision Tools for Farm
Managers." J. Am. Soc. Farm Managers and Rural Appraisers, 51(1987):51-57.

Simons, G.L. Expert Systems and Micros. Manchester, England: The National Computing Centre
Limited, 1985.

Spahr, S.L., and H.B. Puckett. "Electronics and Livestock Production." Illinois Review,
28(1986):22-25.

Weiss, S.M., and C.A. Kulikowski. A Practical Guide to Designing Expert Systems. Totowa, NJ:
Rowman and Allanheld, 1984.

193



194