Elif Kongar*, Mahesh Baral and Tarek Sobh

1
Are We Accepting the Right Students to Graduate
Engineering Programs Measuring the Success of
Accepted Students via Data Envelopment Analysis

• Elif Kongar, Mahesh Baral and Tarek Sobh
• Departments of Technology Management and
  Mechanical Engineering
• University of Bridgeport, Bridgeport, CT, U.S.A
• 2008 ASEE Annual Conference Exposition
• Pittsburgh, PA
• June 22-25, 2008

2
Motivation I Difficulties in admission
procedure due to increasing number of students in
the SOE at UB.
UB SOE Enrollment 2002 - 2008
of Available Dual Degree Programs 16 of
Available Concentration Areas / Graduate
Certificate Programs 34
Being able to admit students in less than 5
minutes
Priceless ?
3
Motivation II
Lack of literature to suggest a solution for
customized curriculum.
Moore (1998) - an operational two-stage expert
system to examine the admission decision process
for applicants to an MBA program, and predict the
degree completion potential for those actually
admitted. Nilsson (1995) - differences in the
predictive relationships between the scores of
the Graduate Record Examination (GRE) and the
graduate grade point average, and the scores of
the Graduate Management Admission Test (GMAT) and
the graduate grade point average. Landrim et al.
(1994) - a value tree diagram for fifty-five
graduate institutions offering the Ph.D. degree
in psychology. The authors used this diagram to
indicate the relative weight of admission factors
used in the decision making process.
4
Introduction Data Envelopment Analysis
Efficiency Output/Input
(year)
(year)
(number)
5
A simple numerical DEA example
Efficiency of Candidate B OB/OV app. 70
6
Two DEA Models

• DEA Model I
• To rank the applicants according to
• e1 number of below-B grades in
  math-related/technical courses in the BS
  transcript of the applicant,
• e2 number of semesters to complete the BS
  degree,
• e3 BS GPA of the applicant,
• e4 TOEFL score of the applicant,
• e5 GRE-Q score of the applicant,
• e6 number of years of work experience of the
  applicant.

7
Two DEA Models

• DEA Model I
• To rank the applicants according to
• e1 number of below-B grades in
  math-related/technical courses in the BS
  transcript of the applicant,
• e2 number of semesters to complete the BS
  degree,
• e3 BS GPA of the applicant,
• e4 TOEFL score of the applicant,
• e5 GRE-Q score of the applicant,
• e6 number of years of work experience of the
  applicant.

8
MS Computer Science Application Data (Fall 2004)
37 Students
Source Office of Admissions, University of
Bridgeport, 2008
9
Relative Efficiency Scores and Ranks of Each
Candidate
10
DEA I - Technical Efficiencies, Min, Mean, Max.
11
Two DEA Models

• DEA Model II
• To rank the applicants according to
• t1 number of below-C grades in the M.S.
  transcript of the M.S. candidate,
• t2 GPA of the M.S. candidate,
• t3 application status for the Curricular
  Practical Training (CPT) or Optional Practical
  Training (OPT).

12
MS Computer Science Application Data (Fall 2004)
t1 number of below-C grades in the M.S.
transcript of the M.S. candidate, t2 GPA of the
M.S. candidate, t3 application status for the
Curricular Practical Training (CPT) or Optional
Practical Training (OPT).
37 Students
Source Office of Admissions, University of
Bridgeport, 2008
13
DEA II - Technical Efficiencies, Min, Mean, Max.
14
Comparing DEA I II Establishing a Pattern
Proposed DEA application detects the efficient
DMU more successfully compared to the ones that
are below the average.
15
(No Transcript)
16
Conclusions
DEA allows introduction of intangible and
out-of-system indicators.
Can accommodate multiple inputs and multiple
outputs.
Allows these inputs and outputs to be expressed
in different units of measurement.
Does not require an assumption of a functional
form relating inputs to outputs.
TE is affected by the performance indicators.
Quality of data is important.
17
Future Research

• Additional criteria
• University ranking
• Problem statement
• Financial statement
• publications/projects
• Quality of publications/projects
• and others
• Weight
• Automated model (DEA Solver Pro v.5.0)
• Database I/O
• Statistics collection
• Predict and compare the degree completion for
  those actually admitted

18
Are We Accepting the Right Students to Graduate
Engineering Programs Measuring the Success of
Accepted Students via Data Envelopment Analysis
Thank you !
Elif Kongar, Mahesh Baral and Tarek
Sobh Departments of Technology Management and
Mechanical Engineering University of Bridgeport,
Bridgeport, CT, U.S.A We would like to
acknowledge the following individuals that
contributed their time and, more importantly,
their innovative ideas to this project. Audrey
Ashton-Savage, Vice President of Enrollment
Management Bryan Gross and Isabella Varga,
Office of Admissions. 2008 ASEE Annual
Conference Exposition Pittsburgh, PA June
22-25, 2007
19
Regression Analysis

• RA A statistical technique used to find
  relationships between variables for the purpose
  of predicting future values.

x1 19.04651 0.02465x2
20
DEA orientation

• Input-oriented DEA models define efficiency as
  the least input for the same amount of output
• Output-oriented DEA models define it as the most
  output for the same amount of input.
• Other considerations
• of DMUs App. 2 to 5 times of the sum of Input
  and Output variables
• Input and output selection

21
Justification of Method Selection

• Data envelopment analysis (DEA) is a widely
  applied linear programming-based technique.
• Low divergence low complexity
• Aim is to evaluate the efficiency of a set of
  decision-making units.
• DEA has mostly been used for benchmarking and for
  performance evaluation purposes.
• A DEA approach to measure the relative efficiency
  of end-of-life management for iron in different
  countries.

22
Advantages of DEA

• Can accommodate multiple inputs and multiple
  outputs
• Allows these inputs and outputs to be expressed
  in different units of measurement.
• It doesn't require an assumption of a functional
  form relating inputs to outputs.
• DMUs are directly compared against a peer or
  combination of peers.
• Efficient units form the efficient frontier and
  inefficient units are enveloped by this frontier
  providing information on their improvement
  potential.

23
Data Envelopment Analysis Model
where, k 1 to s, j 1 to m, i 1 to n, yki
amount of output k produced by DMU i, xji
amount of input j produced by DMU i, vk weight
given to output k, uj weight given to input j.
24
Dual Output-oriented CRS Model
25
Simplified schematic diagram of the application
evaluation and decision making process
26
OCEAN
27
OCEAN Admin Part