| Summary: | 碩士 === 國立成功大學 === 工程科學系專班 === 95 === The current multiple college-entrance education system has brought students much more stress than ever. While the recommendation screening test makes students with special abilities and performances stand out, the unified entrance examination is found to be a lot fairer. In this study, we analyze the influence of students’ entrance examination result of the University of Science and Technology, scholastic performance and attendance of cram schools on their admission to national scientific and technical universities, with the aid of Association Rule, Naïve Bayes Classifier and Decision Tree in Data Mining. Data Mining enables teachers to adopt different enhanced learning strategies to help students with different learning styles, so that students can learn in a more efficient way. Moreover, based on the same database, we also analyze the difference and similarities, advantages and disadvantages between these three Data Mining techniques so as to find out the most appropriate analyze rule dealing with issues concerning supplementary lessons.
The study proves that all the three data mining techniques lead to the result that specialized subjects (circuitry, microprocessor principle, experimental curriculum) and math are the most important factors determining whether a student can be admitted to national scientific and technical universities, and the attendance of cram schools contributes the most. The comparison between the three techniques shows: (1)Association Rule is most suited to be applied to this database with highly correlated scoring data attribution. It is best used in analyzing and inducting recognition science (in the field of education). (2)Naïve Bayes Classifier makes a better prediction concerning the independent features of data’s individual properties. It also makes a quicker and more precise classification. (3)Decision Tree generates more comprehensible rules and clearly interprets important data (the subtree come near the root). It can be applied to almost all studies involving classification, only some difficulties may arise cutting decision tree when there are too many samples.
|
|---|
