Context-Aware Technology Research And Application

In recent years, with the fast development of Pervasive Computing, Electronic Commerce, Social Networking Services and the Mobile Internet, people have been into the era of "Big Data" unconsciously. Information becomes overloaded and information-redundancy becomes one of the top problems of manufacture and academia. Personalized recommender system based on context is the improvement of the traditional recommender system. With the fusion of context, it can analysis the users’idea more accurately and give out a recommendation list with much more accuraey. Therefore, personalized recommender system based on context attracts a great number of researchers. At present, the study of personalized recommender system based on context has just started and there are still many problems to be solved.In this paper, we mainly discuss on some algorithms models about the recommender system and introduce the context awareness after that. The main research contains:the user modeling based on context awareness in the recommender system, the data set of context recommendation, and data correlation algorithm of context, the algorithm of fusion context information into the recommendation and the evaluation of context-awareness recommender system. After the review of context-awareness recommender system, we mainly state our research achievements as below:The friends-recommendation of SNS product (Weibo) based on context. Since SNS is popular recently, it is highly significant to do the research for it. As we all know, the more friends the users have, the more active SNS is. So friends-recommendation for SNS users is also of great importance. In this paper, we imitate the Latent Factor Model, pre-filter the training data with time context and reduce the data noise. We also use the time context into the bias model of users’ and items’(to be-recommended friends) to dig the context of users’activities. The recommendation evaluation turns out to be good with the statistical properties of users’and items’, the extension matrix factorization model and the model validation with the data from Tencent Weibo.The collaborative filtering based on field algorithm. In this paper, we also point out the negative impact the items’ global heat context has on traditional similarity calculation method, i.e. the Pearson similarity formula, and put out the IR-IUF++algorithm which is similar to the thought of TF-IDF. The IR-IUF++algorithm not only reduces the weights of heat items but also finds and raises the weights of items whose evaluation is fluctuate a lot. In this way, we reduce the global influence of heat items. In the experiment, we use Movielens data to conduct the survey and the MAE descent stably.