Cargando...
In this paper the authors describe Pundit, a course recommendation and search tool at Teachers College, Columbia University. The alpha prototype employs a novel combination of data retrieval and data mining approaches to recommend courses to users based on a match between their profiles and course contents. We utilize course management system and library e-reserves data to collect information about course content that is indexed and then matched with user profiles. In conclusion, we define an evaluation function that was used to determine the quality of recommendations.
3 Evaluation. Our evaluation strategy links every user in the system with every course, s/he has taken at Teachers College. We compared this information with recommendations from Pundit to determine the number of intersections. Before doing this, we removed transcripts and/or other course related information from user profiles to eliminate any bias towards the final results. The data used for evaluating the Pundit methodology consisted of 45,713 course files, 3,403 courses, 35,215 student & faculty, 40 user profiles, 74 degree programs, and 247 department. Figure 1 below shows the results we obtained when 200 recommended courses for 40 faculty member profiles were compared against the courses faculty previously taught. Figure. 1. Courses Recommended Vs. Courses Taught.
About this resource...
Visits 159
Categories:
Integrated catalog in space
0 comments
Do you want to comment? Sign up or Sign in