On Validity of Sentiment Analysis Scores and Development of Classification Model for Student-Lecturer Comments Using Weight-based Approach and Deep Learning

Abstract

In this paper, a novel state-of-art classification method was presented for student-lecturer comment classification. Tf-Idf was used to assign weights for each word and several different ANN structures were tested. A large dataset, 52571 comments, was used during training. The results show that developed models clearly overperformed existing classification models in this field. 97% of prediction accuracy was achieved on 3-class dataset, while the prediction accuracy for 5-class dataset was 92%.