Student Enrollment Prediction Model Using Admissions Data: A Data Mining Approach; due June 2008.

(From an abstract submitted July 18,2007:)

In modern world higher education is transitioning from enrollment mode to recruitment mode. This has brought paradigm shift and paved the way for institutional research and policy making from historical data perspective. In recent studies undertaken, researchers have increasingly used neural networks and data mining for performance and persistence prediction as well as dropout problem of students. The primary objective of this research was to build models to predict enrollment using the student admissions data from a four-year public university with enrollment more than 25,000 students. Authors conducted experiments using WEKA, the open source data mining software, and evaluated the models using cross validation, quartile charts, and win-loss tables.

Experimental rig consisted: feature subset selection based on Wrapper and Infogain, and comparison of algorithms (learners), such as, J48, AODE, Naive Bayes, JRip, RIDOR, and OneR using discretization techniques -NBins and Fayyad-Irani. The authors compared results based on accuracy, probability of detection (pd), and probability of false alarm (pf), and evaluated the performance of learners using win-loss tables based on t-tests. To enhance the study, authors combined the features, such as, high school GPA, financial aid indicator, and residency status to form a new goal. Authors conducted similar experiments to compare the performance of the learners, J48 and Naive Bayes for the new goals.

Experiments showed that financial aid was the most important factor to attract students to enroll among other factors, such as, ACT scores, median family income (estimated using zip code), and major, and that a specific class of students can be targeted if some features are combined.

by Ashutosh Nandeshwar (Institutional research information officer), Kent State University, and Subodh Chaudhari (Graduate student), West Virginia University

Combining Complementary Formal Verification Strategies to Improve Performance and Accuracy : defended June 15, 2007

Slides : http://menzies.us/pdf/07owen.pdf

Software : http://unbox.org/wisp/tags/lurch/1.1

Papers:

• The Strangest Thing About Software, January 2007, IEEE Computer.
• Effectively Combining Software Verification Strategies: Understanding Different Assumptions, ISSRE 2006
• more