Application Of Data Mining In Teaching Evaluation System In Colleges And Universities

Issued by the Central Committee of the Communist Party of China and the State Council in October 2020,the "Overall Plan for Deepening Education Evaluation Reform in the New Era" points out that it is necessary to improve results evaluation,strengthen process evaluation,explore appreciation evaluation and promote comprehensive evaluation with adhering to scientific as well as effective methods and that the scientific,professional and objective nature of education evaluation can be improved by utilizing information technology.It demonstrates that,in the new era,the reform of teaching evaluation system is key to cultivating high-quality faculty in the college and university and it is urgent to improve the scientificity and accuracy of the system.This thesis aims to reconstruct the teaching evaluation system by using the teaching evaluation data of college students and related algorithms in data mining so as to establish a more scientific,reasonable and accurate system of teaching quality evaluation,which may contribute to the long-term development of higher education.This thesis employs 151,636 data from a university evaluation center.The main contents are as follows:(1)Sentiment analysis of student evaluation texts based on deep learning.In terms of the sentiment analysis of student evaluation texts,this thesis proposes an improved CNN-Bi LSTM-CA-SA model that introduces a two-layer attention mechanism on the existing CNN-Bi LSTM model—a channel attention mechanism(Channel Attention,CA)in the CNN model,and a sequence attention mechanism(Sequence Attention,SA)in the Bi LSTM model.They are utilized to better capture the word information in the evaluation text,enhance the feature vectors that play an important role in the prediction results but suppress those not related to the model,thus increasing its accuracy and performance.The experiment result shows that compared with the CNN-Bi LSTM model,the model proposed in this thesis enhances the accuracy of model classification,and improves the precision and recall rate.(2)Screening of subject evaluation indicators based on decision tree algorithm.In line with the graded comments,each group is graded and the decision tree algorithm in the classification technology is used to filter the characteristic indicators to further develop the indicator system.This thesis designs a decision tree model of K-means++＿C4.5 in accordance with the characteristics of teaching evaluation data,which better solves the complex and time-consuming calculation of the C4.5 model when dealing with many continuous values.(3)Weight design of evaluation index based on GT-AHP-CRITIC model.With a view to the problem that the traditional indicator system has the same weight and cannot reflect the difference of indicators’ importance,this thesis proposes a subjective and objective weighting model of GT-AHP-CRITIC on the basis of Game Theory(GT)to weight each indicator.The model uses the Analytic Hierarchy Process(AHP)method to calculate the index weights of teachers and experts’ opinions,and then uses the CRITIC method to comprehensively measure the objective weights of the indicators according to their comparative advantage and conflicts.Finally,based on Game Theory(GT),an evaluation model of GT-AHP-CRITIC combination weighting is constructed,which provides a reference for the weight design of teaching evaluation indicators in the college and university.