Browsing by Author "AMNA SHOUKAT"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    AUTOMATIC GENERATION OF TEACHERS’ COURSE PREFERENCES USING DOCUMENT CLUSTERING
    (UMT, Lahore, 2019) AMNA SHOUKAT
    In our higher educational institutes, assignment of courses to the faculty members as per their skills and preferences is a very difficult task. Until now this work has been done either manually by the higher authorities like Deans and Chairpersons of the departments of the universities or by optimization and scheduling algorithms in the literature. In this research work, an initiative towards automated course allocation to the faculty members has been taken on the basis of their skills and course preferences using document clustering algorithm. The automated preferences of courses for faculty have been generated with the help of clustering in which two types of clusters are formed. One type of clusters contains group of teachers having similar skills and courses they have taught in previous years; other type of clusters contains list of similar nature courses. For clustering, the direct clustering method is used from the software package named CLUTO. The dataset of the teachers and of courses have been collected in the form of documents, afterwards pre-processed and visualized to be used further. To get quality clusters for both teachers as well as courses, we have done many experiments on CLUTO by assigning different values to the number of clusters (k). From the results of the experimentation it has been observed that, for k = 16 we have got quality clusters for both teachers and courses documents using direct clustering method. In CLUTO, similarity function cosine fetches smallest possible entropy and largest possible purity against the clusters formed. For the clusters of courses, we’ve got an Entropy 0.297 and Purity 0.667. Similarly against the clusters of the teachers’ documents we’ve achieved an Entropy 0.262 and Purity: 0.656 respectively for k=16. Finally the generated list of preferences of the courses to the faculty members has achieved an average precision 0.721, average recall 0.67 and F measure 0.69.