Hall of Mirrors
Content Analysis of Online Discussion Forums Using Predictive Analytics (304976)
*Urvashi Desai, Miami UniversityVijayalakshmi Ramasamy, Miami University, Oxford
Keywords: predictive analytics, sentiment analysis, graph mining
Online discussion boards designed on an LMS (Learning Management System) are a very effective means for instructors to understand the evolution of learning relationships and their impacts on learning outcomes, especially in large classrooms of undergraduate STEM courses. It provides a platform for students to share classroom experience, their learning, questions, answers, and challenges. Student interactions not only improve their learning experience but also enhances their critical thinking and problem-solving skills. Modeling the discussion board data using network theory and applying predictive approaches including decision tree learning, neural networks (sentiment analysis) and comparing these results with course grades would reveal the impact of the use of a discussion board on performance as a result of peer-interactions and knowledge sharing. Other implications of this analysis could be to find the most challenging/debated topic in a particular course, analyze the leadership and team-based qualities of a group of students, and analyze patterns and trends in female student participation. The predictions are compared to actual grades, and the best fit model will be reported.