Semester

Fall

Date of Graduation

2006

Document Type

Dissertation

Degree Type

EdD

College

College of Education and Human Services

Department

Learning Sciences and Human Development

Committee Chair

Terence C. Ahern.

Abstract

This dissertation examines the roles of learning styles and teaching methodologies within a data mining educational program designed for non-Computer Science undergraduate college students. The experimental design is framed by a discussion of the history and development of data mining and education, as well as a vision for its future.;Data mining is a relatively new discipline which has grown out of the fields of database management and data warehousing, statistics, logic, and decision sciences. Over the course of its approximately 15 year history, data mining has emerged from its genesis within the academic and commercial research and development arenas to become a widely accepted and utilized method of exploratory data analysis for management, strategic planning and decision support. Over the first several years of its development, data mining remained the province of computer scientists and professional statisticians at large corporations and research universities around the world. Beginning in about 1989, these data mining pioneers developed many of data mining's standards and methodologies on large datasets using mainframe computing systems. Throughout the 1990s, as both the hardware and software tools required for the realization of data mining have become increasingly accessible, powerful and affordable, the pool of potential data miners has expanded rapidly. Today, even individuals and small businesses can exploit the power of data mining using freely acquirable open source software packages capable of running on personal computers.;During the growth and development of data mining methodologies however, little research has been dedicated specifically to the pedagogical approaches used in teaching data mining. Educational programs that have evolved have largely remained within Computer Science departments and have often targeted graduate students as an audience. This dissertation seeks to examine the possibility of successful teaching data mining concepts and techniques to a non-Computer Science undergraduate audience. The study approached this research question by delivering a lesson on the data mining topic of Association Rules to 86 participants who are representative of the target audience. These participants were randomly assigned to receive the Association Rules lesson through either a Direct Instruction or a Concept Attainment teaching approach. The students completed Kolb's Learning Styles Inventory, participated in the data mining lesson, and then completed a quiz on the concepts and techniques of Association Rules. A t-test was used to determine if significant differences existed between the scores generated under the two teaching models, and an ANOVA was conducted to identify significant differences between the four learning style groups from Kolb's instrument. In addition to these two statistical tests, the data were also mined using Association Rules and Decision Tree methods.;In both statistical tests, we failed to reject the null hypothesis, finding no significant differences in quiz scores between the two teaching models or among the four learning style groups. Further investigation into the differences among learning styles within teaching models however did reveal that the Assimilator learning style students who received their instruction via Direct Instruction did score significantly higher on the quiz than did their learning style counterparts who received the lesson via Concept Attainment. This finding suggests that although we cannot rely solely on one instructional approach as consistently more effective than the other, there may be instances where the correct instructional choice will positively benefit some learners with certain learning styles. The results of the data mining activities also support this assertion. Association Rules mining yielded no strong relationships between teaching models, learning styles and quiz scores, but Decision Tree mining did reveal a similar pattern of higher scores earned by Assimilator learners within Direct Instruction.;The findings of this study show that effectively teaching data mining concepts to undergraduate non-Computer Science students will not be as simple as choosing one teaching methodology over another or targeting a specific learning style group. Rather, designing instructional activities using teaching methodologies which closely align with predominant learning styles in a classroom should prove more effective. Perhaps the most significant finding of the study is that elementary data mining concepts and techniques can be effectively taught to the target audience. Finally, we recommend that additional teaching methodologies and perhaps different learning style assessments could be tested in the same way as those selected for this study.

Download

Share

COinS