Research On University Evaluation System Based On Support Vector Machines

The evaluation of universities, which is necessary to the supervision and evaluation of the quality of higher education, do have merits. It helps students winnow universities to find the right one they need, guide the financial flow, stimulate the competition between universities, and make people care about the higher education. The evaluation of universities in china began late and improved quickly, though it is not mature enough.Support Vector Machine (SVM) is a new and very promising classification technique. The approach is systematic and properly motivated by statistical learning theory. Training involves separating the classes with a surface that maximizes the margin between them. An interesting property of this approach is that it is an approximate implementation of the Structural Risk Minimization (SRM) induction principle.In this thesis, the theory and method of Support Vector Machines were studied in the given application, the evaluation of universities. After the earnest research about recent advancement and main points of the evaluation process and SVM, we constructed the evaluation system based on SVM.In the process, we proposed three related issues and discussed them respectively. The fist issue was how to evaluate by classifying. We concatenated the vectors of each two schools to be a "big" vector. We could classify such "big" vectors into three types, namely "better", "equal" and "worse", based on what relation between the two schools is. Thus we could tell the relation between any two schools by classifying the "big" vector concatenated from the vectors of them. The second issue was multi-class classification algorithm of Support Vector Machines. The traditional Support Vector Machines only deal with the binary classification. In this paper, we wanted to deal with 3-class classification by one against one method, in which three SVMs were built to distinguish any two classes respectively. The third issue was how to form absolute evaluations based on the results of classification. To provide absolute evaluations, weadopted a round-robin-like mechanism, in which each school was compared with every other school, and received a mark based on the result. Such marks were cumulated to get the final evaluation of the school. We did lots of simulation experiments, and the results showed our system to be superior in performance to those based on Fisher or PCA. We tried different kernels and adjusted parameters to find the best fit for our problem and built our system on it. In the last, we applied our system in emulating adjustment. We changed the weights of some indicators according to the remarks by pedagogues and did some analysis with the results, which could be a basis for further analysis for researchers in the related field.