Research On Educational Administration Data Mining Based On Bayesian K-nearest Neighbour Algorithm And Principal Component Analysis

University educational administration management is an extremely important link in higher education. It is the core and foundation of entire university management. A general view of former teaching management systems shows that most of them are On-Line Analytical Processing (OLAP) system which lack the abilities of comprehensive analysis and assistant decision-making. Furthermore they failed to use the implicit knowledge in the information with over great capacity accumulated during the history. Here the educational administration data mining is implemented by making use of the data mining theory.In this thesis the elementary theory of data mining technology and the main algorithms are discussed firstly. Researching status of data mining is reviewed and summarized. The applications of principal component analysis and Bayesian k nearest neighbor algorithm in data mining are studied. In the thesis principle component analysis is used in synthetical evaluation of the graduating students' achievement. By removing the correlation between the factors the analysis targets are reduced while the amount of information is kept so that the computation is decreased. The thesis adopts principal component analysis and Bayesian k nearest neighbour algorithm in forecast of students' future occupations. The students' achievements are used as feature data and principal component analysis is used to reduce the dimensions of feature data. Bayesian k nearest neighbour algorithm is applied to implement classification, namely prediction of future occupations. Compared with traditional k nearest neighbour algorithm Bayesian k nearest neighbour algorithm is able to determine the parameters of algorithm by Bayesian learning and Markov Chain Monte Carlo algorithm (MCMC).So those parameters need not to be assumed as before and the results of classification are steadier than before. The simulated experiment results prove that a good classification effect can be achieved and adequate occupations for students can be forecasted by combining principal component analysis and Bayesian k nearest neighbour algorithm. A lot of valuable information can be obtained by mining practice in educational administration systems. Such information can improve the synthetical evaluation of the graduating students' achievement, prediction of students' employment direction, and the achievement prediction of the second level examination of computer in this province. It can also promote the development of academy and the cultivation of more various talented persons needed by society.