Research On Recommender System Applied In Open Network Learning Environment

As the old saying goes,"A wise man knows exactly what is happening in the outside world without leaving his home". In the distant past, this sounds like an idiot. However, the rapid development of technology turns people’s lives upside down, and also realize this dream in the modern. Popularity of the network provides the ability to access information anywhere, anytime for everyone. The Open Learning Environment has become a new star in the education system. It appears, so that people can study at home.However. the convenient access to information brought us a series of new problems. People no longer suffer from the lack of information. A new age of information overload is here. So how to get information is no longer the focus of attention. We care more about looking for the useful information from the open learning environment in this vast network of information and knowledge network, This is the biggest problem nowadays.As we all know, to solve the information overload problem,there are two representative programs. One is classified directories and the other is search engines. Categories put some of the most famous websites categorized so that users can find useful things out according to the category. Search engines are working based on keywords entered by the user and give them some related information. But it’s difficult for the search engine to fully meet users’ demand for information. One of the reasons is that the keyword choosed by users played a decisive role in the search process.When a user can not accurately describe their needs, the effect of search engines also greatly reduced. On the other hand, only keyword-based information retrieval are not enough in many cases. And a lot of irrelevant information browsing process will undoubtedly result in the loss of customers. For users who do not explicitly demand, the search engine is powerless. For these customers, the best solution is a person or an automated tool can help filter the information and put forward some suggestions for them to select from. Recommendation system is one of this automated tool. Recommended system matches the preference information in the user model user with feature information in the object model, while using the appropriate recommendation algorithm to calculate, and recommend to the target users information they might be interested in.For open learning environment, the recommendation system is a very important part. It acts based on the history of the learners’ behavior to provide "smart" recommendation service. These recommendations may be some online action, such as reading a message on a conferencing system, doing an exercise, or to take an online examination, or it might just be one of the learning materials. Recommendation of materials in the learning environment is one of the most important applications in recommendation system. Recommender System use multiple users’ opinions to help choosing the material one learner should study next from a large number of potential choices. If the user wants to make progress in certain areas, the recommendation system can also recommend appropriate learning content.It is also helpful in solving the personalization and information overload problem by using material recommendation system in a learning environment.The users’choice always base on some property of the options. Therefore, this paper presents a hybrid recommender system based on the material properties which greatly improving the accuracy and quality of intelligence recommender. This paper:1. describes the history of the development and status of the recommendation system development, and shows its importance in open-learning environment.2. summarizes the main thought of some existing recommendation algorithms and their advantages and disadvantages. Also, this paper discusses the application of these algorithms in an open learning environment.3. presents a mixed attribute-based learning material recommendation system, and introduces genetic algorithm using by this system. The system consists of two modules, one base on the implicit attributes the other is explicit attribute-based. In the first module, we use the weight of implicit properties as chromosome for genetic algorithm. The algorithm is based on the historical rating data to optimize these weights. Then use the nearest neighbor algorithm (NNA) to give recommendations. In the second module, a multidimensional information model is used by introducing the preference vector (PM). After model the learners’interests based on the explicit properties, measure the similarity between the new preference vector and using nearest neighbor algorithm (NNA) to get recommender results. The experimental results show that this method is better than the existing algorithms on the accuracy, and can alleviate the problem of "cold start" and sparsity.