Analysis For Students' Evaluation Of Teaching Information And Research On Scoring Prediction

Educational Data Mining is the study of data generated in education and teaching.The potential value of mining data is counterproductive in teaching practice.Teaching reform has been deepening with the passage of time.Many colleges and universities regard student evaluation activities as one of the important forms and methods to evaluate teachers' teaching quality.At present,there is a system of evaluation in universities,and the data in the system includes ratings and reviews.Most colleges and universities do not use the evaluation system very well,and do not fully explore the information in the evaluation system.The design of the evaluation system does not refer to the students' opinions.Most colleges and universities only perform simple statistical calculations on the score data,and do not explore the student's scoring model.The comment data written by the students has not been systematically processed and is read by the class teacher himself.The heavy workload caused teachers and education administrators to lose patience with the evaluation system,did not seriously analyze the comments,did not grasp the needs of students,and the evaluation system did not really contribute enough strength to the improvement of teaching quality.The research in this paper mainly starts from two aspects:(1)Matrix decomposition and non-negative matrix factorization scoring data.Firstly,the students' scoring mode is visualized,then the students' scoring mode is classified,the data that is meaningless to the evaluation work is removed,and then the matrix decomposition algorithm is used to decompose the scoring data,and the matrix decomposition algorithm is improved to improve the prediction effect of the algorithm..However,there are negative values in the process of decomposition matrix,and negative values are meaningless to the evaluation work.In order to ensure that the elements in the decomposition matrix are meaningful,the author introduces a non-negative matrix decomposition,which is roughly the same as the matrix decomposition,except that it limits the non-negative of the elements in the decomposition matrix.The algorithm guarantees the non-negative of the decomposition matrix,which makes the decomposition matrix result interpretative.The non-negative matrix decomposes the student evaluation data,and the mean square error between the obtained result and the actual evaluation data is 2.96.(2)Research on the text of the commentary,using natural language processing techniques to mine the emotional information in the text.Processing the comment data first classifies each comment,then performs part-of-speech tagging on the results of the word segmentation,and then uses syntactic analysis to analyze the relationship between the words in the comment.The OP(Opinion Parser)extraction algorithm is used to extract the Aspect words and the Opinion words in the comment.The extracted aspect pairs map the word vector model and the cosine similarity to the six dimensions most relevant to the educator and the teacher,and calculate the specific scores of the students' evaluation of the teacher in each dimension.The precision of the OP extraction algorithm is 81.14%,and the recall is 82.28%.The accuracy of mapping the Aspect words to the six dimensions using the word vector model and the cosine similarity algorithm is 82%,and the author's improvement of the word vector mapping improves the accuracy by 7%.Using the CNKI dictionary to classify the extracted Opinion words,the accuracy is 94.12%.Finally,students are predicted to evaluate teachers in multiple dimensions.Predicting students' evaluation of teachers can help college administrators better understand teachers,make teachers better understand their own advantages and disadvantages,improve teaching,promote teaching progress more smoothly,and improve the quality of teaching in colleges and universities.