Abstract:
|
When presented in a business class setting, the concepts and practices of data mining can often present complex challenges as time considerations and course sequencing prevent full discussions of prevailing mathematical principles lying behind each analysis. Students struggle with appropriate analysis choice. Over the course of four semesters, data has been collected on multiple classes considering the use of real, messy, and varying data sets which immerse students in hands-on problem solving. In the first iteration one "real" data set was given and three contrived. Each semester moving forward an additional "real" data set replaced a current existing textbook data set. Generally, each student team uses a different approach. Yet, when applied properly (with logic and fundamental statistical assumptions), similar results are obtained. The number of clear, actionable conclusions given with understandable visualization added measurable value. The study initiated using schooling survey data from Bolivia, allowing students to use data mining techniques to identify the most impoverished students in high schools in different regions of the country. Additional data sources were used, including entrepreneurial Amazon inventory data with branding and marketing issues; an electric and light company optimizing fan displays across multiple venues while minimizing inventory; and also a supply chain company looking to get information from large troves of available data. Immersing the students in real-world problems and encouraging them to clean and understand the data before applying data mining tools and visualization allows them to become confident problem solvers, a challenge faced in traditional testing. Students clearly value these initiatives.
|