Study On The Application Of Neural Network In The Evaluation Of University Websites

With the development of higher education computerize, information technology is widely used in universities, the typical application are the university websites. University websites play important roles in promoting management recycling, accelerating educational reform, and improving educational resource sharing, facilitating interaction between teachers and students, and enhancing university social influence. There are some problems in commonly used evaluation methods of university portal, such as too many human factors may impact the objectivity of evaluation, evaluation tools are not accurate and lack of uniform evaluation criteria. Therefore, establish an advanced university websites evaluation model, and improve existing evaluation standards, and make objective and accurate evaluation is very important. It also has very important practical significance and practical value.Artificial neural network as the simulation of human brain structure and function has associative memory and self-learning function.BP neural network as the essence of artificial neural network, easier to handle in parallel, non-linear problems, with adaptive capabilities.BP neural network fully display a strong advantage on researching and treating complex issues.Through natural, non-linear network modeling process, without distinction between deliberately linear relationship, it also can continue to learn from a large number of complex data and find the rule.Since the last century it has been proposed, artificial neural network has attracted the attention of psychologists, and other scientists in computer science, information science and continue developing. Present it has a wide range of applications in artificial intelligence, data mining, speech recognition, system simulation, system identification, computer image processing and industrial control.Commonly used methods to evaluating the webite by expert,mainly rely on the experience and ability of the person, so it's very subjective.And the human brain can not handle relationship between multiple targets in parallel,results will deviate largely,it is difficult to ensure objective and fair.To decrease the possible outcome differences caused by subjectivity of evaluators.The paper use BP artificial neural network to construct the artificial neural network model,and improve the traditional evaluation process.Let the evaluation system can automatically adjust to objective environments and continuously improve the accuracy of the evaluation.In order to prepare this paper, I learn from previous experts with reality, and my work could be summarized as follows:1. Generally introduce research results at home and abroad and ordinary methods of evaluation of university websites. Introduce some commonly used qualitative,quantitative and comprehensive evaluation methods on website evaluation,include:URL evaluation, form evaluation, link analysis, software testing method, automatic evaluation, website log file data mining, neural network concepts and steps of evaluation, AHP,information architecture evaluation and fuzzy comprehensive evaluation;Briefly introduce the development and the history of artificial neural networks.Analysis learning methods of the BP neural network and draw out the flow chart.Describe the problems on traditional algorithm such as needing long time to calucate, converge slowly, network may not converge to the minimum, etc.Useing improved activation function and the additional momentum of the Gauss-Newton algorithm for model design.2. By analyzing the existing website evaluation index system, sum up the experience of their predecessors, taking into account the actual situation of high schools and indicators of objectivity, logic, highlighting the importance of the content, the principles of comprehensiveness, set up a new set of website evaluation index system, and draw the index system model diagram. Through useing AHP, calculate the weight of each index and synthetic weight, design questionnaires and the pre-survey questionnaire proves it effective. Provided ways to qualitative and normalized qualitative indicators of index system, so that they can be used as input data of neural network.3. Analysis the probability of artificial neural network; introduce the principles and methods by which improving the current evaluation of university websites with artificial neural network. Design the evaluation model network structure,including: network layers, network input layers and output layers, the hidden layers, expect error, learning rate and the datapreprocessing methods. Constructed the evaluation model based on artificial neural network by use matlab software; Collect evaluation data of thirty universities in Anhui Province,use twenty samples to train the network,adapt the weights and threshold of network,and use the other ten samples as test data.Through analysis of test results prove that the network model is effective, it's performance meet to the requirements fully.Design graphical user interface by using matcom tools and VC++ mixed programming.Application N fold cross-validation method for network performance testing.Analysis the advantages of neural network method compared to traditional methods. Mention the difficulties during the research and the disadvantages,such as network learning cost too long time,samples are deficient, propose ideas to further improve the research.