[ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ 13 ] [ 14 ] [ 15 ] [ 16 ]

Journal of Information Science and Engineering, Vol. 37 No. 5, pp. 993-1009

Exploiting Machine Learning and Feature Selection Algorithms to Predict Instructor Performance in Higher Education

Electronics and Computer Discipline
Indian Institute of Technology, Roorkee
Saharanpur Campus, Saharanpur, 247001 Inida
E-mail: fahujaravinder022; scs60fptg@gmail.com

Machine learning has emerged as the most important and widely used tool in resolving the administrative and other educational related problems. Most of the research in the educational field centers on demonstrating the student’s potential rather than focusing on faculty quality. In this paper the performance of the instructor is evaluated through feedback collected from students in the questionnaire form. The unlabelled dataset is taken from UCI machine learning repository consisting of 5820 records with 33 attributes. Firstly, the dataset is labelled(three labels) using agglomerative clustering and the k-means algorithms. Further, five feature selection techniques (Random Forest,Principal Component Analysis, Recursive Feature Selection, Univariate Feature Selection, and Genetic Algorithm) are applied to extract essential features. After feature selection, twelve classification algorithms (K Nearest Neighbor, XGBoost, Multi-Layer Perceptron, AdaBoost, Random Forest, Logistic Regression, Decision Tree, Bagging, LightGBM, Support Vector Machine, Extra Tree and Na¨ıve Bayes) are applied using Python language. Out of all algorithms applied, Support Vector Machine with PCA feature selection technique has given the highest accuracy value 99.66%, recall value 99.66%, precision value 99.67%, and f -score value 99.67%. To prove that results are statistically different, we have applied ANOVA one way test.

Keywords: classification algorithms, decision tree, ensemble learning, faculty performance evaluation, feature selection, support vector machine

  Retrieve PDF document (JISE_202105_01.pdf)