Research Of Outlier Analysis With Clustering Algorithm For Teacher Appraise

Educational evaluation and activities are important parts in the educational process. The activities of evaluation in education process in a reasonable manner is the guarantee to improve the quality of education, Since no evaluation leads to blind education. Teacher evaluation is an integral part of educational evaluation process. Teacher evaluation is usually with questionnaire. These questionnaires are often with false answer, such as malicious intention and so on. These data will leads the results of the evaluation become unreal, it will also affect the enthusiasm of teachers, even accumulated comprehensive assessment of teachers. Therefore, the exclusion of these data is absolutely necessary.At present, most schools take no action to these issues, or simply to remove the maximum and minimum factors. This simple approach is not scientific.With above problems, this article attempts to remove these abnormal data points, clean the result, accurate data sets with clustering data mining methods. In this paper, we use density-based clustering algorithm to identify outliers, and then use anomaly local detection algorithm (LOF) on the clustering results for further analysis of the degree of abnormality, and finally determine the ultimate collection of abnormal points in the actual situation.In this paper, we complete the following tasks:1. in-depth study of density-based DBSCAN clustering algorithm with improved algorithm IDBSCAN. Through the experimental comparison of the two and found that the improved efficiency of the latter one.2. to find out a collection of outliers LOF algorithm with an analysis of their abnormalities, abnormal points and the final set. In this paper, the calculation of outlier factor with the method mentioned in the literature reflects not only the concept of local anomaly, but also reducing the computational complexity.