A Research On Learning Process Evaluation Based On Support Vector Machine

Learning assessment plays a very important role in the classroom session, since teachers can understand the performance of students quickly in the classroom. Traditional evaluation methods are used to examine student learning outcomes by the exam results in the era of examination-oriented education, but they can’t meet the development of society in the era of quality education. Therefore, a new evaluation method is produced, namely process evaluation. Not only does it promote students’ academic progress, but it optimizes teaching. But there are still flaws in the process of evaluation. For example, it can easily lead to reducing of reliability and validity of process evaluation to teachers’ prejudice for learners, logic errors and inadequate information gathering; and it is to increase teachers’ workload that process evaluation is used to learn. So it is necessary to make a more scientific approach to solve the problems of process evaluation.Data mining has been applied to teaching quality evaluation, distance education and formative evaluation of online learning and so on, because it can handle potentially useful information about massive data, and get good value for education. By analyzing the results of process evaluation, it can be seen that it is a classification problem. Classification is a important task in data mining, so data mining methods are used to process evaluation. Compared with other classification algorithms, Support Vector Machines(SVMs) have many unique advantages, such as solving the small sample problems, nonlinear and high dimensional pattern recognition problems. Thus, SVMs algorithms are chosen to predict the final outcome of the evaluation process in the paper.Taking into account the limitations of process evaluation, SVMs algorithms are used to analyze it. The main work of the paper is as follows:(1) Evaluation scale is chosen to implement process evaluation in the classroom through analyzing three forms of process evaluation and according to the specific needs. It is built to evaluation system about “C Programming” and appropriate rating scale based on the evaluation index system, according to the basic principles of building process evaluation index system.(2) SVMs are applied to the learning process of the evaluation, because the content of evaluation may be missing and unfairness, and the nonlinear relationship between evaluation indexs. The model of process evaluation is constructed, and it is used to solve problems based on SVMs.(3) It is implemented to process evaluation teaching in “C Programming” course of freshman students in Department Education and Science, Datong University, then rating scale is filled out. Then multiple data sets are obtained by collecting the data of the scale evaluation. So LIBSVM is used to do simulation experiments for data sets by using MATLAB software. The results show that SVMs have a good prediction accuracy for the unlabeled samples in multiple data sets. Compared with prediction accuracy of SVMs and BP neural network, the results show that prediction accuracy of OAO is better than BP network to classify multiple data sets. OAO can be good predict unlabeled samples.