TY - GEN
T1 - Can we predict student learning performance from LMS data? A classification approach
AU - Dutt, Ashish
AU - Ismail, Maizatul Akmar
PY - 2019/6
Y1 - 2019/6
N2 - The Learning Management System (LMS) is a common occurrence in most educational institutions. This system is a software application helping the educator in administration, facilitation, and tracking of course content to the learner. Educators have always been interested in understanding student interaction with systems like LMS. Such a system generates a plethora of data in a various form such as student performance on the individual course, activities, student behaviors, etc. The most prominent solutions involve performing dimensionality reduction technique to improve classifier accuracy and reducing the fewer error rates. Therefore, this study utilizes feature selection as a dimensionality reduction technique. The multiclass data were handled using the Learning Vector Quantization (LVQ) algorithm to identify significant predictors and thereby reducing the biased result. The efficiency of feature selection technique is evaluated with five different classifiers such as Linear Discriminate Analysis (LDA), Classification and Regression Tree (CART), k-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest (RF). The performance of the classifier is evaluated using the kappa statistics and confusion matrix. Our extensive experimental results show that RF classifier produces optimum kappa statistic (85%) with LVQ
AB - The Learning Management System (LMS) is a common occurrence in most educational institutions. This system is a software application helping the educator in administration, facilitation, and tracking of course content to the learner. Educators have always been interested in understanding student interaction with systems like LMS. Such a system generates a plethora of data in a various form such as student performance on the individual course, activities, student behaviors, etc. The most prominent solutions involve performing dimensionality reduction technique to improve classifier accuracy and reducing the fewer error rates. Therefore, this study utilizes feature selection as a dimensionality reduction technique. The multiclass data were handled using the Learning Vector Quantization (LVQ) algorithm to identify significant predictors and thereby reducing the biased result. The efficiency of feature selection technique is evaluated with five different classifiers such as Linear Discriminate Analysis (LDA), Classification and Regression Tree (CART), k-Nearest Neighbor (KNN), Support Vector Machine (SVM), and Random Forest (RF). The performance of the classifier is evaluated using the kappa statistics and confusion matrix. Our extensive experimental results show that RF classifier produces optimum kappa statistic (85%) with LVQ
U2 - 10.2991/iccie-18.2019.5
DO - 10.2991/iccie-18.2019.5
M3 - Conference Paper
T3 - Advances in Social Science, Education and Humanities Research
SP - 24
EP - 29
BT - Proceedings of the 3rd International Conference on Current Issues in Education (ICCIE 2018)
A2 - Mustadi, Ali
PB - Atlantis Press
CY - Dordrecht Netherlands
ER -