Master Dissertations

Permanent URI for this collection

http://192.168.18.118:4000/handle/20.500.12661/349

Browse

Browsing Master Dissertations by Subject "Academic problems"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Performance prediction in mathematics using educational data mining techniques: a case of Mzumbe University in Tanzania
    (The University of Dodoma, 2020) Mushi, Paul Kavishe
    Nowadays, Higher Learning Institutions (HLIs) store a large amount of students’ data. However, those data are not widely used to solve the academic problems of the students that are available at the HLIs such as poor performance of the students in some of the courses. Educational Data Mining (EDM) is the technology that can be applied to predict the performance of the students on the available dataset at the HLIs. This study intended to solve the problem of poor performance in Mathematics for degree management students at HLIs using EDM techniques. The purpose of the study was to predict Management degree Students’ performance in Mathematics using EDM taking Mzumbe University (MU) as a case study. The quantitative research approach was applied in this study basing on the design science steps. Secondary data were collected to create the dataset through document review from examination office (final examination (FE), course work (CW) and Remarks), admission office (age, gender, entry category and ordinary level mathematics grades), accounts office (sponsorship details), department of mathematics and statistics (number of instructors) and accommodation office (living location) at MU including Main campus Morogoro and Mbeya Campus. Different Machine Learning (ML) algorithms were applied on training set (60%) such as K-Nearest Neighbor (K-NN), Random Forest (RF), Decision Tree (DT), Support Vector Classification (SVC) and Multilayer Perceptron (MLP). ML algorithms were validated using 10-fold cross-validation and validation dataset (20%) and the best algorithms were RF, DT and K-NN. During the evaluation of the three best ML algorithms using 20% of the dataset, RF ML algorithm was found to be the best for model development in mathematics performance prediction in this study with the accuracy of 99% and F1-scores of 99% and 100% for fail and pass class respectively. Moreover, DT was able to generate rules that were applied to recommend the minimum grade of D for ordinary level mathematics in admission to degree management students to reduce the failure rate at HLIs.