Felege, christopher online education perceptions and recommendations focus ...
Kishant Chatarpal Poster - SURF 2013
1. A Data-driven Empirical Model of Choices –
Illustrated by Undergraduate Course Selection and GPA
Motivation and Background
Much of what we know about the choices
people make is based on information gathered
in contrived settings wherein people make one
choice at a time. We know little about how
people make multiple choices over time in
situations where a variety of factors are at
play together, and what the consequences are
of these choices. It will be of value to theory,
and in practice, for us to understand these
choices and consequences in real situations.
Today’s technologies allow us to capture large
amounts of data on actual choices that people
make. The goal of this research is to develop
an empirical driven-model of choices and
consequences by using what are called data
mining techniques on such data. The focus is
on students’ choices of courses at UH and
their GPAs.
Data were compiled on 377 students (177Jrs;
& 200 Srs), who were pre-business majors as
of Spring 2013. The data also include their
gender, race, courses completed, GPAs, the
professors who taught them, and their
academic levels over time. A measure of
semester by semester GPA progress was
derived, with values of “G” when GPA
increased or remained the same and “B” when
GPA decreased, for consecutive semesters.
The choices of interest are the courses, and
the professors selected. The consequences of
interest are the cumulative GPA and the
progressive GPA.
Two data mining techniques are used. One is
Affinity Analysis, which is used to identify
patterns of choices and expresses these
patterns in “if choice A, then choice B” rules.
We use this technique to derive the choice
models for courses and professors selected.
The other is a Classification and Regression
Tree, which is used to group outcomes based
on their “similarity in outcomes” and with
respect to other given factors.
Research Approach
• Students do not appear to choose courses in ways that help to
increase their GPAs (See Tables above and Fig 3).
• Students choose courses and professors independently (See Tables).
• As a result of students’ choices, there are distinct groups of courses
that yield higher GPAs (See Fig 3).
• When course choices result in better GPAs, this happens at the
Sophomore, Junior, and Senior levels (See last graph in Fig 3).
• As a result of their choices, students make similar and better kinds of
GPA progress at the Sophomore and Senior levels (See Fig 2).
• Similar and better progress in GPA is made by Female students, who
choose certain courses (See last graph in Fig 2).
• As a result of their choices, white students have similar and higher
GPAs than Hispanic, Asian and Black students (See Fig 1)
• Similar and higher GPAs are earned by female white students, who
choose certain courses and as Seniors (See last graph Fig 1).
Conclusion
Results
By: Kishant Chatarpal | Faculty Mentor: Dr. Norm Johnson
Fig 1: Similarity in GPA predicted by Race, Gender, Courses & Acad. Level
Fig 2: Similarity in GPA Progress predicted by Gender, Courses & Acad. Level
Fig 3: Similarity in GPA predicted by Courses and Acad. Level only
I am grateful for the support I received from: Dr. Johnson, Dipti
Katkar, Todd Chaykosky, Karen Weber, SURF Honors Program,
Bauer College, and the Provost Office.
Acknowledgements
The following graphs show the consequences of students’ choices.
Model of Choices Made by Students