Content clustering of computer mediated courseware using data mining technique

Abstract

Computer Mediated Courseware (CMC) has been developed so far for individual courses considering single or multiple text books. A group of courseware can be developed by using multiple text books and in this case, it is a requirement to cluster the contents of different books to form a generalized clustered content. No work has been found to develop this generalized clustered content. In this thesis, we have proposed a methodology based on data mining techniques to construct a hierarchical general structure of a group of courseware combining the individual structure of a set of books. We have represented each book by a tree structure where the root of the tree is the book and the nodes are section, subsection etc. The leaf nodes are keywords set for section or subsection. For clustering purpose, we have transformed each of the book trees into relational database format and applied data mining clustering algorithms on these data. We have considered title of sections, subsections, synonym and homonym of keywords during the clustering process. For experiment, three standard text books on database have been used. We have measured the performance of our algorithm considering different level of keyword set selected by domain expert. We have achieved more than 95% true positive and less than 5% false positive. The clustering will help the courseware developer to dynamically allocate contents to develop different courses using a group of books. We have applied this methodology for different level of courses on database; the methodology is generalized and can be applied to any other courses.