Quality Evaluation of Dissertation Based on DBSCAN Clustering

The quality of dissertations is an important reference for evaluating the quality of graduate degree-awarding. Most of the methods evaluate the quality of the dissertation from four types of dimentions, topic and review, innovation and paper value, scientific research ability and basic knowledge, and the standard expression of the paper. However, the experts' comments often have great value for evaluating the quality of the dissertations. Thus, this paper proposes an innovative method of quality evaluation of the dissertations, which integrates the experts' comments and four types of evaluation criteria mentioned above. The results show that the deviation of the dissertations' score becomes larger. So, it is easier to distinguish between excellent papers and problematic papers, as the score of low-level papers decrease, while the score of high-level papers increase. And, our proposed method has a strong generalization ability for all levels of the paper. Finally, the paper compares the indicators and comments of the eliminated papers and excellent papers in the engineering discipline, analyzes the typical characteristics of the two kinds, and gives corresponding suggestions on how to write high-level dissertations.