ABSTRACT
Instructor evaluation based on student feedback is essential in education, it allows instructors to see if their teaching has been effective. But, it is very challenging for an instructor who teaches numerous students to analyse feedback provided by all the students. To solve this problem, this paper develops a sentiment analysis model to analyse students’ feedback to assess the effectiveness of teaching and learning. In this paper, machine learning models like, Support Vector Machines, Multinomial Naive Bayes, Random Forests, K-Nearest Neighbours and Neural Networks are trained on feature engineering and re-sampling techniques to classify student feedback into three sentiment classes: negative, positive or neutral, using student dataset collected from Kaggle. From the analysis before the resampling of the data, K-Nearest Neighbours model is found to be more efficient in predicting student sentiment towards teaching practices than the other models with good accuracy of 81%. After the resampling of the data, Neural Networks performed better than the other models with good accuracy of 84%. The model will help institutions make effective decisions towards teaching and learning strategies.
