Analysis Of Students' Evaluation Teaching Data Based On Multi-dimensional Outlier Detection

With the continuous development of teaching reform,schools pay more and more attention to the assessment of teachers' teaching quality.As the most direct participants and beneficiaries of teaching activities,students' evaluation of teaching has become an important form and means of teaching quality evaluation and are widely used in the education industry.In recent years,most colleges and universities in the teaching evaluation activities mainly adopt the online evaluation method.The Academic Affairs Office and the Teaching Evaluation Center will comprehensively evaluate the teaching quality of teachers at the end of the semester or the end of the year,The evaluation data is weighted and statistically by the evaluation system,and the final teaching quality ranking is obtained.Supervise and improve the teaching of the bottom 5% teachers,so as to improve the overall teaching level of the school.However,a total score is usually obtained after weighted statistical analysis.This method has a single evaluation standard,and many "problem data" will be drowned out in the calculation of weighted and average,and can't fully reflect the real situation of a teacher in teaching activities.Aiming at the traditional teaching evaluation method can't realize the deep mining of hidden "problem data",this paper proposes a multi-dimensional outlier detection teaching quality evaluation method.Firstly,compared with the scores obtained by weighted statistics,the initial evaluation index data can better reflect the teacher's teaching performance in various aspects.After statistical analysis,this paper uses the form of finding the variance to one-way detect outliers.Secondly,as the most direct result of teaching activities,the students' scores can reflect the quality of teaching more directly.the correlation coefficient between student achievement and each evaluation index is calculated,and student achievement is taken as an important parameter for binary outlier detection.Finally,both the scores and the students' grades reflect the quality of teaching through abstract values.Comparatively speaking,the comments given by students in the process of teaching evaluation can directly reflect the students' emotions about teaching activities.calculates the emotional extreme value of students' comments and adds it to the ternary outlier detection of teaching quality as an important indicator.Through the experiment of the evaluation results of a university in 2010,it is concluded that the outlier detection method adopted in this paper can not only find thesamples which rank lower in the traditional teacher assessment,but also find some samples which can not be found in the traditional teacher assessment.Therefore,it is verified that the teaching evaluation method in this paper is more effective and which provides a more perfect scientific basis for the teaching quality evaluation system.