Variance Analysis In Teaching Evaluation Themes Of Students With Different Majors Based On LDA Model

In universities,students’ evaluation of teaching is an important activity to ensure the quality of teaching.The data of teaching evaluation are comprehensively analyzed to get the students’ satisfaction level to the teachers of the classes,which provides some reference value for the subsequent teaching reform.However,most universities do not reflect the special characteristics of different majors in the objective index setting of the teaching evaluation questionnaire.In fact,if the objective questions of the questionnaire are designed in categories,more accurate and effective feedback information will be collected and the effectiveness of teaching evaluation will be improved.In this thesis,we address this issue and conduct a study based on the subjective evaluation text data of students from different majors.A study was conducted to analyze the differences in the focus of the fine-grained thematic distribution of the evaluation text data among different majors.The quantitative analysis and qualitative research were used to determine which aspects students in different majors focus more on in the learning process of the course.Overall,the main elements and innovations of this thesis are as follows.(1)The differences in the fine-grained thematic distribution of the textual data of students’ evaluation in different majors are discussed in terms of different teaching methods,both online and offline.The study shows that foreign language and literature majors pay more attention to the linguistic presentation and the breadth of knowledge of the teacher,statistics majors pay more attention to the organization and logic of the teacher,and computer science majors pay more attention to whether the teacher has enough experience and depth of expertise.(2)To address the question of whether the teaching quality is "homogeneous" when the teaching modes are different.This study analyzes and compares the evaluation text data of offline teaching mode with the valid evaluation text data of undergraduate courses in an online teaching semester in a university in Jiangxi.The results show that there is a significant difference between the online and offline evaluation results of computer science students among the three types of majors.This may be due to the fact that there are more hands-on courses in this category,and the practical courses cannot achieve good teaching effect when they are taught online.(3)Text-based topic feature extraction.In this thesis,we propose a Word2vec-LabeledLDA-KNN model with LDA model as the main framework for topic clustering analysis of evaluation texts.The model introduces the Word2 vec model based on the Labeled-LDA model to solve the short text feature sparsity problem,and the KNN model is added to reduce the misclassification rate when classifying topics.Compared with the previous model,the accuracy of the improved model reaches 91.58% and the recall rate reaches 92.62%,which is better than the original LDA model and labeled-LDA model,and the value of clustering entropy of the new model is about 0.04 to 0.13 lower than the other models.To sum up,the subjective teaching evaluation texts of students are mined in multiple aspects from the real-life problems.The findings of this thesis can also provide some reference value for such universities like a university in Jiangxi in setting objective indicators of teaching evaluation questionnaires and subsequent teaching reforms.