Computer Science and Systems Analysis

Computer Science and Systems Analysis

Technical Reports
Miami University Year 

The Graduate Student Advisor (GSA):

An Expert System for SAN Graduate
Student Advising
Jiazhu Zhang
Miami University, [email protected]

This paper is posted at Scholarly Commons at Miami University.

http://sc.lib.muohio.edu/csa techreports/43
 DEPARTMENT OF COMPUTER SCIENCE
& SYSTEMS ANALYSIS

TECHNICAL REPORT: MU-SEAS-CSA-1992-016

The Graduate Student Advisor (GSA): An Expert

System for SAN Graduate Student Advising
Jiazhu Zhang

School of Engineering & Applied Science | Oxford, Ohio 45056 | 513-529-5928

 The Graduate Student Advisor (GSA):
An Expert System for SAN
Graduate Student Advising

Jiazhu Zhang
Systems Analysis Department
Miami University
Oxford, Ohio 45056

Working Paper #92-016 December, 1992

 T h e G r a d u a t e S t u d e n t A d v i s o r (GSA):
An E x p e r t System for SAN G r a d u a t e Student A d v i s i n g

Jiazhu Zhang

Department of Systems Analysis

Miami University

Abstract

The Graduate Student Advisor (GSA) is an experimental expert system

that advises graduate students in systems analysis. It simulates a
faculty advisor in suggesting the schedule of courses a student
should take based on the student's background and interests. It is
implemented in NEXPERT Object. This report first describes the
task, knowledge engineering and solution approach of the GSA. The
report then gives a sample session to illustrate how to run a
consultation. It also includes some maintenance notes about how to
modify the knowledge about courses and related rules if necessary.
Finally, it discusses some lessons learned.

1. Introduction

The Systems Analysis Master's Degree Program at Miami offers a

blend of computer science, information systems and operations
research. The program curriculum consists of four elements:
Foundation Courses, Core Courses, Elective Courses and Graduate
Research Courses. A graduate student must complete at least 4
courses(l2 credit hours) from each of Core and Elective elements
and at least 2 graduate research courses (6 credit hours).
Successful completion of a minimum of 30 credit hours is required
for the degree. Additional foundation courses may be necessary
depending on the student's different interests and background.

Currently, there are approximately 40 graduate students enrolled in

the Systems Analysis Graduate Program. The number is expected to
be increasing in the coming academic year. Due to the conversion
nature of the program, the students come from a variety of
backgrounds. Some may have backgrounds in a systems-related field
such as systems analysis or computer science. Some may have
backgrounds in a non-systems field such as geology or geography.
There are many other different possibilities in between these two
extreme cases. According to the program, students must have had:
1 communication course, 2 calculus courses, 2 probability and
statistics courses, and 1 computer programming course. Some
students may have to make up some of the undergraduate deficiency
courses before they take any core and/or elective courses. Also,
most students pursue their graduate study on a full-time basis
though there are some part-time students. Students are supported
through various financial means (e.g. graduate assistantship,
grant-in-aid or scholarship from an organization). The minimum
number of credit hours for which a full-time student must register
and the maximum number of credit hours for which the student may
register in a regular term vary depending on the financial support
means. All of these make advising students even more complicated.

Advising students is a complex and time consuming task. However,

a large percentage of an adviser's time is spent with fairly
repetitive activities. For example, the process for determining
course prerequisites and planning programs of study may seem
confusing to the individual student. The Graduate Student Adviser
(GSA) is an experimental expert system that simulates a faculty
adviser in suggesting the schedule of study a student should take
for obtaining the degree. Its purpose is to provide students with
the curricular knowledge and information on a uniform and
consistent basis, thus making more efficient use of the graduate
adviser's time.
2. Knowledge Structure and Components

GSAts approach is largely data driven; it begins with a set of

"best" courses and tries to produce a schedule within the
constraints imposed by the properties of the courses, relationships
between them, and restrictions on the schedule.

The knowledge for GSA has been elicited from several experts
including SAN graduate faculty members, professors of Mathematics
and Statistics Department, and the Associate Dean of the Graduate
School. GSA contains two kinds of knowledge:
* knowledge about courses -- course number, title, credit
hours, prerequisites, offering time and so on;
* knowledge about constraints -- rules for determining course
prerequisites, the "bestw courses and rules for determining
the schedule length and forming the schedule of study.
It is implemented in NEXPERT Object. The program currently has
about 550 rules and 50 objects (and classes). Knowledge about
courses is stored as objects (and/or classes) . An object is the
fundamental unit of knowledge representation in NEXPERT Object. A
given object may be defined to belong to one or more classes, which
determine the names and types of its properties. It may include
subobjects which are in turn full-fledged objects with properties
and subobjects of their own. A class is merely a grouping of a set
of objects. The class definition may include any number of
properties to be inherited by the objects belonging to the class
(referred to as its instances or members). The class may have any
number of subclasses, which will likewise inherit its properties
and pass them on in turn on to their own instances, A given class
may be a subclass of more than one other class (called
superclasses), just as a given object may be an instance of more
than one class or a subobject of more than one other object.
Consider the following example, which describes the course SAN572.
Name SAN572
Classes Core-Courses
 San-Grad-Courses
...
subobjects . . .
Properties c r e d i t - h o u r s : 3
p r e r e q u i s i t e s : Unknown
d e s c l : Systems life cycle; problem definition; gathering information;
desc2 : creative problem-solving; project management; feasibility study;
desc3 : alternative technology selection; system requirements;structured systems
desc4 : analysis tools; data flow diagrams and dictionary, algorithms
d e s c 5 : specification; logical design; user procedures; review of translation of
d e s c 6 : logical design into systems design. Normally students work on systems
desc7 : development project.
no-of -d e s c s : 7
t i t l e : Analysis of Information Systems
weight : Unknown
when-o f f e r e d : fall, spring

Below is another example, which describes the category of Core

Courses.
Name Core-Courses
Subclasses . . .
properties c r e d i t -h o u r s : Unknown
p r e r e q u i s i t e s : Unknown
d e s c l : Unknown
desc2: Unknown
d e s c 3 : Unknown
d e s c 4 : Unknown
desc5: Unknown
d e s c 6 : Unknown
d e s c 7 : Unknown
d e s c 8 : Unknown
d e s c 9 : Unknown
no-of -d e s c s : Unknown
t i t l e : Unknown
we i g h t : Unknown
when-o f f e r e d : Unknown

Knowledge about constraints is represented by rules. A rule is the

basic unit of inference and reasoning in NEXPERT Object. Every
rule has three basic parts:
*One or more conditions;
 *Exactly one hypothesis;
*Zero or more actions.
The following shows an English translation of a sample GSA rule:
IF the student is a full time student
and
the student is supported through a graduate assistantship
THEN setting the rnax and min numbers of total term credit hours
ACTIONS set the max number of total term credit hours to 16;
set the min number of total term grad credit hours to 10.

Objects (and classes) in NEXPERT Object are essentially the same as

frames suggested in an A 1 setting by Minsky [ 5 ] . Frames are
formalized structures for representing knowledge. They are often
linked together in a network, thus effectively handling the major
inadequacy of production rules for defining terms and for
describing domain objects and relationships among objects [I, 3 1 .
A frame commonly consists of two parts: a name and a set of
attribute-value pairs. It provides a structured representation of
an object, or a class of objects. The use of frames increases the
efficiency of the stored information processing by attaching
procedures to nodes which know how to compute values of variables
in response to queries and how to update values of variables in
response to assertions [ 3 ] . However, the procedures attached to
some slots of individual frames are insufficient to organize the
whole computation. The rule representations play the major role in
organizing the whole computation.

Objects represent the knowledge being reasoned on by the rules.

Hierarchical relationships can be defined between objects to give
rules greater reasoning flexibility over objects. The rules,
objects (and classes) and relationships between them form the whole
knowledge base. The rule and object (and class) relations in
NEXPERT Object are illustrated in Figure 1. Note that the
relations can be static or dynamic. For example, the relations
between courses and their prerequisites are established according
to the student's background and exist only for the duration of the
advising session in which they are created.

A Rule A Class An Object A Slot

Figure 1. Rule and Object Relations in NEXPERT Object

3. Constructive Problem Solving

The distinguishing feature of constructive problem solving is that

solutions have to be constructed rather than selected [ 3 ] . Typical
6
tasks that require constructive problem solving are planning,
design and certain kinds of diagnoses [ 3 3 . In each of the cases,
it is infeasible to fix the solution set in advance, There are
many different ways in which actions can be ordered, components can
be assembled, and faults can co-occur.

Our student advising problem is essentially a constructive problem.

Students come from different backgrounds and have different
interests. There are too many possible programs of study in which
courses can be planned.

Advising a graduate student can be considered as a multi-phase

process [ 7 , 83 :
(a) determine how many courses the student should take;
(b) determine which courses the student may take, based on
prerequisite requirements and the student's academic history;
(c) determine the ''bestw courses based on the student's
interests;
(d) generate a program of study from the best courses.
The first three phases determine the needs of the student, and are
thus of a diagnostic character, while the last phase plans the
student's schedule.

The determination of the "best1' courses is based on the course

weights. One course is said to be the "bestw if it has the highest
weight. Each course has a preassigned individual weight to each
interest area. The weight of a course is computed by summing up
the preassigned individual weights of the course to the interest
areas of the student (See the Appendix A for the course individual
weights).

The last phase is the main task: construction of the program of

study. It can be divided into three subtasks:
(1) determine the schedule length;
(2) plan SAN790 courses;
(3) generate each semester's schedule from the best courses.
In the context of the last subtask, there are several constraints
used to form each semester's schedule. Here are some examples:
* Each course must be offered in the current semester.
* The prerequisites of each course must be taken or planned
prior to the current semester.
* The total number of credit hours planned'in the current
semester must not exceed the maximum number of credit hours a
full time student may register in a regular semester.

Although GSA may have several attempts at planning a course, it

never backtracks. In other words, it never makes a decision which
it later has to go back and undo. At any point in the problem-
solving process, it has enough knowledge to recognize what to do;
this cuts down on trial-and-error scheduling. Backtracking is
computationally expensive, especially in terms of run-time [ 3 ] .

4. Running An Advisory Session

As described above, the main task of GSA is advising the student

and drawing up a program of study based on the student's background
and interests. Before the system begins to make any inferences for
the case at hand, the student will be asked a series of questions.
There are about 25 questions which may be put to the student.
Which of these questions will be posed vary from case to case, In
other words, a fixed set of questions is not asked in each case;
rather the questions asked depend on answers to previous questions.
All the questions are posed in its natural way. If a question
requires an answer other than a simple "yes/noW answer, a pop-up
list, choice list or selection table is attached to the question.
For some questions, there will be a help information pop-up
available.

As a result of the dialogue, GSA starts planning the student's

schedule. When construction of the student's schedule is
completed, the schedule will be presented to the student, The
student could print out the schedule or look at the course
descriptions. After the student looks over the schedule, he/she
could decide to accept the suggestion or make any change under
course constraints and restrictions on the schedule. The whole
process is very straightforward.

The following is a self-explanatory example of a session with GSA

(See the Appendix B for running the GSA). What would actually
appear on the screen uses pop-up windows. It is very difficult to
draw those pop-up windows here. Therefore, the format has been
changed for the purpose of illustration.

--GSA: On what basis do you intend to pursue your graduate study?

Full-time or part-time?
USER: full-time
--GSA: How are you going to be supported financially in your graduate study?
Graduate assistantship;
Modified graduate assistantship;
Graduate grant-in-aid;
other support means.
USER: graduate assistantship
--GSA: Please indicate the first semester of the study schedule you would like
to let GSA create for you.
USER: fall, 1992
--GSA: Is this going to be your first semester in the SAN graduate program at
Miami?
USER: y
--GSA: Have you ever taken any systems analysis classes (at the 500 or higher
level) and/or statistics classes (sta563, sta663) at Miami before?
USER: n
--GSA: The following is a list of current SAN graduate faculty research
interests. Please indicate which areas you would be interested in:
Research Areas Graduate Faculty
 Artificial Intelligence M. Ozden, A. Sanders
Database Systems Y. Uckan
Software Engineering J. Kiper
...
Undecided
USER: Database Systems, Software Engineering
--GSA: What is your undergraduate major?
Computer Science;
Decision Science;
Management Information System;
Mathematics;
Operations Research;
Statistics;
Systems Analysis;
Other.
USER: Mathematics
--GSA: Have you ever taken any communication class before?
USER: n
--GSA: Please choose one of the following suggested communication classes:
com135 Essentials of Public Speaking;
com138 Technical Communication;
com231 Small Group Communication.
USER: com135
--GSA: How many probability and statistics classes have you taken before?
USER: 2
--GSA: Have you ever taken any class(es) in Engineering Economics, Linear
Programming, and/or Stochastic Process?
USER: y
-GSA: Are you familiar with any programming language?
USER: y
--GSA: Do you have any background in Data and File Structures?
USER: n
--GSA: Do you have any background in Computer Architecture and Assembly Language
Programming?
USER: n
--GSA: Based on the information you have provided about your background and
interests, GSA suggests to you the following schedule:
Fall. 1992 Total Credit Hours: 13
coml35 3 Essentials of Public Speaking
san586 3 Artificial Intelligence
san603 3 Program, Data and File Structures
san604 3 Computer Organization and Operating Systems

Spring. 1993 Total Credit Hours: 12

san571 3 Simulation
san572 3 Analysis of Information Systems
san623 3 Data Communication, Networks and Distributed Processing
san624 3 Expert Systems

Fall. 1993 Total Credit Hours: 1 2

san573 3 Automata, Formal Languages, and Computabiliy
san575 3 Structured Design and Implementation
san612 3 Data Structures and Database Systems
san790 3 Graduate Research

Spring, 1994 Total Credit Hours: 12

san574 3 Language, Interface, and Their Processors
san622 3 Advanced Database Systems
sta663 3 Introduction to Applied Probability
san790 3 Graduate Research

Would you like to make any changes ? (Y/N)

USER: y
--GSA: To which semester would you like to make changes?
USER: Spring, 1994
--GSA: You may check any courses you would not want to take in Spring, 1994:
-
CRHR COURSE TITLE
san574 3 Language, Interface and Their Processors
san622 3 Advanced Database Systems
*sta663 3 Introduction to Applied Probability
 san790 3 Graduate Research
You may choose the courses you would like to take in Spring, 1994:
-
CRHR SEMESTER PREREOUISITES COURSE TITLE
san582 3 Fall none Analysis of Inventory Systems
*san584 3 Spring none Analysis of Manufacturing Systems
...
--GSA: On the basis of the information you have provided about your background
and interests, GSA suggests to you the following schedule:
Fall. 1992 Total Credit Hours: 13
com135 3 Essentials of Public Speaking
san586 3 Artificial Intelligence
san603 3 Program, Data and File Structures
san604 3 Computer Organization and Operating Systems

Spring.
- 1993 Total Credit Hours: 12
san571 3 Simulation
san572 3 Analysis of Information Systems
san623 3 Data Communication, Networks and Distributed Processing
san624 3 Expert Systems

Fall. 1993 Total Credit Hours: 12

san573 3 Automata, Formal Languages, and Computabiliy
san575 3 Structured Design and Implementation
san612 3 Data Structures and Database Systems
san790 3 Graduate Research

Spring. 1994 Total Credit Hours: 12

san574 3 Language, Interface, and Their Processors
san622 3 Advanced Database Systems
san584 3 Analysis of Manufacturing Systems
san790 3 Graduate Research

Would you like to make any changes ? (Y/N)

USER: n
5. Maintainerws Manual

This section describes how to modify the knowledge about courses

and the related rules if necessary. It is assumed that you know
NEXPERT Object.
5.1 Update Offering Time of a Course

First, use Object Editor to find the object with the course number
as object name. Then activate Meta-slot Editor to update the value
of the property when-offered.

5.2 Update the Description of a Course

Updating the course description is similar to updating the course

offering time. First, use Object Editor to find the object with
the course number as object name. Note that there are 9 properties
descl, ... , desc9, which can be used to define up to 9 lines of
course description. Each line can hold about 70 characters.
Activate Meta-slot Editor to modify the values of properties descl,
..., desc9. Then, use Meta-slot Editor to update the value of
property no-of-descs, which hold the actual number of lines of the
course description.

5.3 Add A New Course

Currently, GSA has the knowledge about all SAN graduate courses.
In the case that a new course should be offered in the future time,
you can add it to the knowledge base in NEXPERT Object. First, use
Object Editor to create an object with the course number (e.g.
san626) as the object name. Fill in the class field of the Object
Editor window one of the class names (Foundations, Cores,
Electives) to indicate that the course is a foundation course, core
course, or elective course. The object created will have the same
properties as other SAN courses belonging to the same class have.
Then use Meta-slot Editor to define the values of the properties:
course number, course title, credit hours, offering time, course
description, prerequisites. There are nine properties: descl, .,.,
desc9, which can be used to define the course description. Thus,
you can define up to 9 lines of course description. Put the actual
number of lines in property no-of-descs. If the course does not
have any prerequisites, assign "nonew to property prerequisites.
When it has prerequisites or is prerequisite of other courses, you
need to update those rules dealing with course prerequisites. Here
is what needs to be done. If the course has foundation course(s)
as its prerequisites, use Rule Editor or Rule Notebook to find
.
those rules with hypothesis "set-grad-course~rereq'~ Then use
Rule Editor to modify the action part of those rules. For example,
say, san603 is a prerequisite of the new course san626. Add to the
action part of those rules with condition "1s student,san603-ok
Ilnflwthe following three actions:
Do 11san6031@ san626 .prerequisites
createobject san603 san626
Do MAX(san603.wgt, san626,wgt+l) san603,wgt
If the new course is a prerequisite of other SAN course(s), use
Rule Editor or Rule Notebook to find the rule with hypothesis
t8setgrereq_wgttt. Then use Rule Editor to modify it. For example,
say, the new course san567 is a prerequisite of san654. Add the
statement:
CreateObject san567 san654
to the condition part and the following two actions:
Do 1gsan567e1 san654 ,prerequisites
Do MAX(san567.wgt, san654.wgt+l) san567,wgt
to the action part of the rule. If the new course has as its
prerequisites some course(s) other than foundation courses, similar
modifications should be made to the rule with hypothesis
wset-prereq-wgt". Finally, you have to update the weight list of
each interest area. First, determine the weight of the new course
to each interest area. Then, Add the weight of the new course to
the end of the wgt-list property value of the corresponding
interest area. Find the rule with hypothesis I t i n i t-dialogue88.
Then, add the new course number (e.g, san567) to the end of the
String Value in the second condition.

6. Lessons Learned

(1) One of the major pitfalls to be avoided in developing an expert

.
system is choosing an inappropriate problem [2, 3, 4, 93 Here are
some important criteria relevant to the selection of an appropriate
problem for expert system development [2, 3, 4, 9, 101:
*The application task must have a well-defined domain;
*One or more experts must have the knowledge required;
*Those experts must be able to verbalize desired task
performance;
*The task does not depend heavily on common sense;
*The task is of managementable size.
Not picking the "rightw problem can lead to complications, even
failure in the subsequent development of the system.

(2) There is a fundamental difference between an expert system

shell and a conventional programming language: shells are object
oriented and knowledge intensive, while conventional programming
languages are procedure oriented and code intensive [63. However,
it is not necessary that programming in shells should be any easier
to debug and result in less effort. In conventional programming,
there exist notions of what constitutes good programming practice
[3]. Such is less the case in knowledge engineering [3]. My own
experience of programming in NEXPERT Object suggests that the level
of programming skill required by a shell should not be
underestimated.

(3) Every aspect of the advising process has to be mapped out in

detail, and every alternative has to be explored to the point of a
conclusion. The expert system can start with only the intelligence
that is put into it. If that information is confusing or
incomplete, then your "expertw will be also. A large part of the
total effort in creating an expert system must take place before
you even touch the keys of a computer.

(4) The process of building an expert system is inherently

experimental [3]. It is reported that simple expert systems have
been built in as little as 3 man-months [2]. The time for expert
system development with present techniques appears to be around 5
man-years per system [2]. If you see a demonstration of a really
good expert system that took three scientists years to develop
using Prolog, Lisp or some expert language on a mainframe, do not
expect to create a similar system in weeks on a microcomputer by
using an expert system shell.

7, Conclusion

The process of building the Graduate Student Advisor was very

educational and insightful. The GSA program performs well. The
decomposition of student advising process into phases has been of
help in the following ways: (1) it simplifies interactions with the
domain experts; (2) it reduces the complexity; (3) it provides
effective modularization. The GSA approach is not necessarily
limited to advising graduate students. It could be applied in
solving some other problems such as personal financial planning
[71
References

1. Fikes, R. and Kehler, T. (1985). The Role of Frame-based

Representation in Reasoninq. Communication of the ACM, September,
pp. 904-920
.
2. Gevarter, W. B. (1990) The Basic Principles of Expert Systems.
In Raeth, P. G. Expert Systems: A Software Methodoloqy for Modern
Applications. Los Alamitos, California: IEEE Computer Society
Press. pp. 17-32
.
3. Jackson, P. (1990) Introduction to Expert Systems 2nd edn.
Wokingham UK: Addison-Wesley
4. Liebowitz, J. and De Salvo, D. A. eds. (1989). Structurinq
Expert Systems: Domain, Desisn, and Development. Englewood Cliffs
NJ: Yourdon Press
5. Minsky, M. (1975). A Framework for Representins Knowledqe. In
Winston, P. ed. The Psycholoqy of Computer Vision. New York:
McGraw-Hill. pp. 211-277
6. Raeth, P. G. (1990). Two PC-based Expert System Shells for the
First-time Developer. In Raeth, P. G. Expert Systems: A Software
Methodoloqy for Modern Applications. Los Alamitos, California: IEEE
Computer Society Press. pp. 2-6
7. Valtorta, M. G., Smith, B. T. and Loveland, D. W. (1984). The
Graduate Course Advisor: A Multi-phase Rule-based Expert System.
Report No. CS-1984-18, Dept. of Computer Science, Duke University
8. Valtorta, M. G. (1983). The Graduate Course Adviser. Master
Project Report, Dept. of Computer Science, Duke University
9. Waterman, D. A. (1986). A Guide to Expert Systems. Reading,
Massachusetts: Addison-Wesley
10. Williams, C. (1990). Expert Systems, Knowledqe Enqineerinq, and
A 1 Tools: An Overview. In Raeth, P. G. Expert Systems: A Software
Methodoloqy for Modern Applications. Los Alamitos, California: IEEE
Computer Society Press. pp. 2-6
 Appendix: Survey Results on the Course Weights

Note: I n t h e survey, SAN g r a d u a t e f a c u l t y members w e r e a s k e d t o

e v a l u a t e t h e i n d i v i d u a l weight of each S A N g r a d u a t e c o u r s e t o e a c h
r e s e a r c h i n t e r e s t a r e a u s i n g a s c a l e from 0 t o 5. E i g h t of them
responded t o t h e survey.

SAN571 SAN572 SAN573 SAN574 SAN575 SANS82 SAN583 SAN584

Algorithms 1 1 5 3 2 1 1 1

Artificial Intelligence 1 1 3 3 1 1 1 1

Computer Assisted Instruction 3 3 2 3 3 0 0 0

Computer Graphics 2 1 1 1 1 0 0 0

Data Corranunication & Computer 2 2 2 2 1 0 0 0

Networks

Database Systems 2 2 3 3 2 0 0 0

Expert Systems 2 3 2 3 2 1 1 1

Forecast Systems 3 2 1 1 1 3 5 3

Information Retrieval 2 4 2 2 1 0 0 0

Inventory systems 3 2 0 0 0 5 4 3

Machine Learning 2 0 4 3 1 0 0 0

Mathematical Optimization 3 1 2 1 1 3 2 3

Natural Language Processing 1 2 5 5 2 1 0 0

Operating Systems 2 1 3 3 1 0 0 0

Programming Languages 1 1 4 5 3 0 0 0

Queueing Systems 4 0 0 0 0 2 2 2

Simulation 5 2 0 0 1 3 2 2

Software Engineering 2 5 4 4 5 0 0 0

Systems Dynamics 1 3 0 0 0 2 2 1

Theory of Computation 1 1 5 4 2 0 0 0
(continued)
SAN586 SANG01 SAN602 SAN603 SAN604 SANG12 SAN613 SAN614

Algorithms 3 5 4 5 1 3 4 1

Artificial Intelligence 5 3 3 3 2 4 1 1

Computer Assisted Instruction 3 3 1 3 2 3 1 3

Computer Graphics 1 3 1 3 2 1 1 1

Data Comunication & Computer 1 3 3 3 3 2 2 2

Networks

Database Systems 3 3 1 5 3 5 1 1

Expert Systems 5 4 3 4 2 3 2 1

Forecasting Systems 1 2 4 1 0 0 2 2

Informational Retrieval 2 3 1 3 3 4 1 1

Inventory Systems 1 2 5 2 0 1 4 3

Machine Learning 5 3 2 3 3 2 1 1

Mathematical Optimization 3 4 5 2 2 1 5 3

Natural Language Processing 5 3 2 2 2 2 1 1

Operating Systems 1 4 1 4 5 3 1 1

Programming Languages 3 3 1 5 5 3 0 0

Queueing Systems 1 3 5 3 2 2 3 4

Simulation 2 5 5 3 2 1 3 5

Software Engineering 2 4 1 5 5 4 0 0

Systems Dynamics 2 2 3 1 0 0 2 4

Theory of Computation 2 3 2 2 3 2 2 1
Alnori thms
--

Artificial Intelligence
- -

Computer Assisted Instruction

--
1 2

Computer Graphics

Data Communication & Computer 2 2 5 1 4 1 1

Networks

Database Systems 1 5 3 3 3 0 1

Exwert Systems 2 2 1 5 2 1 1
--
Forecasting Systems
- -
lo
Information Retrieval
- -

Inventory Systems

Machine Learning

Mathematical Optimization 1 0 1 1 1 3 11
Natural Language Processing 1 2 I 2 I 2 I 4 I1 I1 I 1 II
Operating Systems I 4 1 5 I 1 II
Programming Languages 3 1 2 2 2

Queueing Systems 1 0 2 1 1

Simulation 1 0 2 2 1

Software Engineering 5 3 3 2 3

Systems Dynamics 0 0 0 2 0

Theory of Computation 1 1 0 1 1
 Appendix B: Running the GSA

GSA RTD.
The GSA system consists of the following files (total 77 files):
GSA. TKB
FRM1. FRM
.0 .

FRM69. FRM
COURSE.TXT
CRSEDESCOTXT
SCHEDULE. TXT
SEMESTER. TXT
WAIT. TXT
HLP2 0 HLP
The file GSA.RTD is the Runtime Definition File which is used to
start the GSA. The file GSA.TKB is the knowledge base developed
under NEXPERT Object. The files with .FRM extension are the
NEXPERT forms which are used to create customized interfaces.
COURSE.TXT, CRSEDEC.TXT, SCHEDULE.TXT and SEMESTER.TXT are NEXPERT
report files used to display course descriptions and schedules.
WAIT.TXT and HLP2O.HLP are just DOS text files used to display some
information to the users.
Suppose that the directory \GSA contains all the files of GSA. In
order to start NEXPERT Forms with GSA automatically, NXPFORMS.EXE
should also be under the same directory \GSA. To start GSA on your
system from the directory \GSA, type
NXPFORHS /FGSA.RTD
at the prompt. This command automatically starts PJEXPERT Forms
with GSA.
If NXPFORMS.EXE is not under the directory, you have to start
NEXPERT Forms environment by typing the command
NXPFORMS
at the prompt. Then use the System menu to read the runtime
definition file of GSA and start GSA.