An Evaluation Method for Teachers Based on the Quantitative Fuzzy Decision Optimization

Admittedly, it is of paramount guiding significance to construct a scientific comprehensive evaluation system for university teachers' teaching performance to promote the overall improvement of university teaching quality. Despite the continuous emergence of evaluation models for teachers' performance, they are typically beset with two significant challenges, one of which lies in their limited ability to describe the expression degree of prior knowledge based on the importance degree of features and to optimize the fuzzification process quantitatively. The other challenge can be summarized as the failure of consequent parameters of fuzzy rules in executing the solution process conveniently and adaptively, as well as the adverse influence of redundant information in fuzzy rules on the model's prediction results. To tackle the foregoing challenges, this paper proposes a new fuzzy classification strategy to determine the corresponding important factors by leveraging the importance of critical features selected based on information gain. This is further employed to realize quantitative optimization of the features' membership degree, thereby improving the overall fuzzification process. On the other hand, this paper proposes a novel method for obtaining the consequent parameters, which can not only optimize the redundant expression in fuzzy rules by using multiple correlation coefficients but also flexibly make the predicted output infinitely close to the true value, thus improving the generalization performance of the model and enhancing its linguistic interpretability. Also, this aligns with the explainable artificial intelligence (XAI), a current research hotspot. The experimental results demonstrate that the model proposed in this paper exhibits excellent decision-making ability, generalization performance, and linguistic interpretability on the adopted teacher performance evaluation datasets.